Update ort and improve the speed of preprocessing

* Add onnx proto

* Update ort to 2.0.0-rc.2

* Improve the speed of resizing

* Fix yolo-seg bug

* Update README.md

Cargo.toml

@@ -12,17 +12,18 @@ exclude = ["assets/*", "examples/*"]
 [dependencies]
 clap = { version = "4.2.4", features = ["derive"] }
 ndarray = { version = "0.15.6" }
-ort = { version = "2.0.0-alpha.4", default-features = false, features = [
+ort = { version = "2.0.0-rc.2", default-features = false, features = [
     "load-dynamic",
     "copy-dylibs",
-    "profiling",
     "half",
     "ndarray",
     "cuda",
     "tensorrt",
     "coreml",
-    "ureq",
     "openvino",
+    "rocm",
+    "openvino",
+    "operator-libraries"
 ] }
 anyhow = { version = "1.0.75" }
 regex = { version = "1.5.4" }
@@ -41,3 +42,6 @@ image = "0.25.1"
 imageproc = { version = "0.24" }
 ab_glyph = "0.2.23"
 geo = "0.28.0"
+prost = "0.12.4"
+human_bytes = "0.4.3"
+fast_image_resize = "3.0.4"

README.md

@@ -4,9 +4,10 @@ A Rust library integrated with **ONNXRuntime**, providing a collection of **Comp
 
 ## Recently Updated
 
-|        Portrait Matting         |
+
+|        YOLOv8-Obb         |
 | :----------------------------: |
-|<img src='examples/modnet/demo.png'   width="800px">|
+|<img src='examples/yolov8/demo-obb-2.png'   width="800px">|
 
 
 |        Depth-Anything         |
@@ -14,17 +15,16 @@ A Rust library integrated with **ONNXRuntime**, providing a collection of **Comp
 |<img src='examples/depth-anything/demo.png'   width="800px">|
 
 
+|        Portrait Matting         |
+| :----------------------------: |
+|<img src='examples/modnet/demo.png'   width="800px">|
+
+
 |        YOLOP-v2          |             Face-Parsing             |               Text-Detection           |  
 | :----------------------------: | :------------------------------: |  :------------------------------: |
 |<img src='examples/yolop/demo.png'  height="180px">| <img src='examples/face-parsing/demo.png'  height="180px"> | <img src='examples/db/demo.png'  height="180px"> |
 
 
-|        YOLOv8-Obb         |
-| :----------------------------: |
-|<img src='examples/yolov8/demo-obb-2.png'   width="800px">|
-
-
-
 
 
 ## Supported Models
@@ -94,7 +94,7 @@ check **[ort guide](https://ort.pyke.io/setup/linking)**
 
 ## Integrate into your own project
 <details close>
-<summary>Check Here</summary>
+<summary>Expand</summary>
 
 #### 1. Add `usls` as a dependency to your project's `Cargo.toml`
 
@@ -107,7 +107,7 @@ cargo add --git https://github.com/jamjamjon/usls
 ```Rust
 let options = Options::default()
     .with_model("../models/yolov8m-seg-dyn-f16.onnx");
-let mut model = YOLO::new(&options)?;
+let mut model = YOLO::new(options)?;
 ```
 
 - If you want to run your model with TensorRT or CoreML
@@ -129,7 +129,7 @@ let mut model = YOLO::new(&options)?;
 
   ```Rust
   let options = Options::default()
-      .with_confs(&[0.4, 0.15]) // person: 0.4, others: 0.15
+      .with_confs(&[0.4, 0.15]) // class 0: 0.4, others: 0.15
   ```
 - Go check [Options](src/options.rs) for more model options.
 

examples/clip/main.rs

@@ -15,7 +15,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_profile(false);
 
     // build model
-    let model = Clip::new(options_visual, options_textual)?;
+    let mut model = Clip::new(options_visual, options_textual)?;
 
     // texts
     let texts = vec![

examples/db/main.rs

@@ -6,18 +6,18 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i00((1, 4, 8).into())
         .with_i02((608, 960, 1280).into())
         .with_i03((608, 960, 1280).into())
+        // .with_trt(0)
         .with_confs(&[0.4])
         .with_min_width(5.0)
         .with_min_height(12.0)
-        // .with_trt(0)
         .with_model("ppocr-v4-db-dyn.onnx")?;
 
-    let mut model = DB::new(&options)?;
+    let mut model = DB::new(options)?;
 
     // load image
     let x = vec![
         DataLoader::try_read("./assets/db.png")?,
-        // DataLoader::try_read("./assets/2.jpg")?,
+        DataLoader::try_read("./assets/2.jpg")?,
     ];
 
     // run

examples/depth-anything/main.rs

@@ -7,7 +7,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i00((1, 1, 8).into())
         .with_i02((384, 512, 1024).into())
         .with_i03((384, 512, 1024).into());
-    let model = DepthAnything::new(&options)?;
+    let mut model = DepthAnything::new(options)?;
 
     // load
     let x = vec![DataLoader::try_read("./assets/2.jpg")?];

examples/dinov2/main.rs

@@ -7,7 +7,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i00((1, 1, 1).into())
         .with_i02((224, 224, 224).into())
         .with_i03((224, 224, 224).into());
-    let _model = Dinov2::new(&options)?;
+    let _model = Dinov2::new(options)?;
     println!("TODO...");
 
     // query from vector

examples/face-parsing/main.rs

@@ -10,7 +10,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         // .with_trt(0)
         // .with_fp16(true)
         .with_confs(&[0.5]);
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/nini.png")?];

examples/fastsam/main.rs

@@ -8,7 +8,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i02((416, 640, 800).into())
         .with_i03((416, 640, 800).into())
         .with_confs(&[0.4]);
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/bus.jpg")?];

examples/modnet/main.rs

@@ -7,10 +7,10 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i00((1, 1, 4).into())
         .with_i02((416, 512, 800).into())
         .with_i03((416, 512, 800).into());
-    let model = MODNet::new(&options)?;
+    let mut model = MODNet::new(options)?;
 
     // load image
-    let x = vec![DataLoader::try_read("./assets/portrait.jpg")?];
+    let x = vec![DataLoader::try_read("./assets/liuyifei.png")?];
 
     // run
     let y = model.run(&x)?;

examples/rtdetr/main.rs

@@ -4,9 +4,9 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
     // build model
     let options = Options::default()
         .with_model("rtdetr-l-f16.onnx")?
-        .with_confs(&[0.4, 0.15]) // person: 0.4, others: 0.15
+        .with_confs(&[0.4, 0.15])
         .with_names(&coco::NAMES_80);
-    let mut model = RTDETR::new(&options)?;
+    let mut model = RTDETR::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/bus.jpg")?];

examples/rtmo/main.rs

@@ -8,7 +8,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_nk(17)
         .with_confs(&[0.3])
         .with_kconfs(&[0.5]);
-    let mut model = RTMO::new(&options)?;
+    let mut model = RTMO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/bus.jpg")?];

examples/svtr/main.rs

@@ -8,7 +8,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_confs(&[0.2])
         .with_vocab("ppocr_rec_vocab.txt")?
         .with_model("ppocr-v4-svtr-ch-dyn.onnx")?;
-    let mut model = SVTR::new(&options)?;
+    let mut model = SVTR::new(options)?;
 
     // load images
     let dl = DataLoader::default()

examples/yolo-world/main.rs

@@ -7,9 +7,9 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i00((1, 1, 4).into())
         .with_i02((416, 640, 800).into())
         .with_i03((416, 640, 800).into())
-        .with_confs(&[0.3]) // shoes: 0.2
+        .with_confs(&[0.3])
         .with_profile(false);
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/bus.jpg")?];

examples/yolop/main.rs

@@ -6,7 +6,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_model("yolopv2-dyn-480x800.onnx")?
         .with_i00((1, 1, 8).into())
         .with_confs(&[0.3]);
-    let mut model = YOLOPv2::new(&options)?;
+    let mut model = YOLOPv2::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/car.jpg")?];

examples/yolov5/main.rs

@@ -10,13 +10,12 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_anchors_first(true)
         .with_yolo_task(YOLOTask::Segment)
         .with_model("yolov5s-seg.onnx")?
-        .with_trt(0)
-        .with_fp16(true)
+        // .with_trt(0)
+        // .with_fp16(true)
         .with_i00((1, 1, 4).into())
         .with_i02((224, 640, 800).into())
-        .with_i03((224, 640, 800).into())
-        .with_dry_run(3);
-    let mut model = YOLO::new(&options)?;
+        .with_i03((224, 640, 800).into());
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/bus.jpg")?];

examples/yolov8-face/main.rs

@@ -8,7 +8,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i02((416, 640, 800).into())
         .with_i03((416, 640, 800).into())
         .with_confs(&[0.15]);
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/kids.jpg")?];

examples/yolov8-falldown/main.rs

@@ -3,7 +3,7 @@ use usls::{models::YOLO, Annotator, DataLoader, Options};
 fn main() -> Result<(), Box<dyn std::error::Error>> {
     // build model
     let options = Options::default().with_model("yolov8-falldown-f16.onnx")?;
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/falldown.jpg")?];

examples/yolov8-head/main.rs

@@ -3,7 +3,7 @@ use usls::{models::YOLO, Annotator, DataLoader, Options};
 fn main() -> Result<(), Box<dyn std::error::Error>> {
     // build model
     let options = Options::default().with_model("yolov8-head-f16.onnx")?;
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/kids.jpg")?];

examples/yolov8-trash/main.rs

@@ -5,7 +5,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
     let options = Options::default()
         .with_model("yolov8-plastic-bag-f16.onnx")?
         .with_names(&["trash"]);
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/trash.jpg")?];

examples/yolov8/main.rs

@@ -3,21 +3,25 @@ use usls::{coco, models::YOLO, Annotator, DataLoader, Options};
 fn main() -> Result<(), Box<dyn std::error::Error>> {
     // build model
     let options = Options::default()
-        .with_model("yolov8m-dyn.onnx")?
+        // .with_model("yolov8m-dyn.onnx")?
+        // .with_model("yolov8m-dyn-f16.onnx")?
         // .with_model("yolov8m-pose-dyn.onnx")?
         // .with_model("yolov8m-cls-dyn.onnx")?
-        // .with_model("yolov8m-seg-dyn.onnx")?
+        .with_model("yolov8m-seg-dyn.onnx")?
         // .with_model("yolov8m-obb-dyn.onnx")?
         // .with_model("yolov8m-oiv7-dyn.onnx")?
         // .with_trt(0)
         // .with_fp16(true)
+        // .with_coreml(0)
+        // .with_cuda(3)
         .with_i00((1, 1, 4).into())
         .with_i02((224, 640, 800).into())
         .with_i03((224, 640, 800).into())
-        .with_confs(&[0.4, 0.15]) // person: 0.4, others: 0.15
+        .with_confs(&[0.4, 0.15]) // class 0: 0.4, others: 0.15
         .with_names2(&coco::KEYPOINTS_NAMES_17)
+        // .with_dry_run(10)
         .with_profile(false);
-    let mut model = YOLO::new(&options)?;
+    let mut model = YOLO::new(options)?;
 
     // build dataloader
     let dl = DataLoader::default()

examples/yolov9/main.rs

@@ -7,8 +7,8 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .with_i00((1, 1, 4).into())
         .with_i02((416, 640, 800).into())
         .with_i03((416, 640, 800).into())
-        .with_confs(&[0.4, 0.15]); // person: 0.4, others: 0.15
-    let mut model = YOLO::new(&options)?;
+        .with_confs(&[0.4, 0.15]);
+    let mut model = YOLO::new(options)?;
 
     // load image
     let x = vec![DataLoader::try_read("./assets/bus.jpg")?];

src/core/annotator.rs

@@ -495,7 +495,7 @@ impl Annotator {
 
                 // keypoint
                 let color = match &self.keypoints_palette {
-                    None => self.get_color(i + 10),
+                    None => self.get_color(i),
                     Some(keypoints_palette) => keypoints_palette[i],
                 };
                 imageproc::drawing::draw_filled_circle_mut(

src/core/dataloader.rs

@@ -81,7 +81,7 @@ impl DataLoader {
         let n_new = paths.len();
         self.paths.append(&mut paths);
         println!(
-            "{CHECK_MARK} {n_new} files found ({} total)",
+            "{CHECK_MARK} Found images x{n_new} ({} total)",
             self.paths.len()
         );
         Ok(Self {

src/core/device.rs

@@ -1,5 +1,6 @@
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord)]
 pub enum Device {
+    Auto(usize), // TODO
     Cpu(usize),
     Cuda(usize),
     Trt(usize),

src/core/engine.rs

@@ -1,12 +1,25 @@
 use anyhow::Result;
 use half::f16;
+use human_bytes::human_bytes;
 use ndarray::{Array, IxDyn};
 use ort::{
-    ExecutionProvider, ExecutionProviderDispatch, Session, SessionBuilder, TensorElementType,
-    TensorRTExecutionProvider, ValueType,
+    ExecutionProvider, Session, SessionBuilder, TensorElementType, TensorRTExecutionProvider,
+    MINOR_VERSION,
+};
+use prost::Message;
+use std::collections::HashSet;
+
+use crate::{
+    home_dir, onnx, ops::make_divisible, Device, MinOptMax, Options, Ts, CHECK_MARK, CROSS_MARK,
 };
 
-use crate::{home_dir, Device, MinOptMax, Options, CHECK_MARK, CROSS_MARK, SAFE_CROSS_MARK};
+/// Ort Tensor Attrs: name, data_type, dims
+#[derive(Debug)]
+pub struct OrtTensorAttr {
+    pub names: Vec<String>,
+    pub dtypes: Vec<ort::TensorElementType>,
+    pub dimss: Vec<Vec<isize>>,
+}
 
 /// ONNXRuntime Backend
 #[derive(Debug)]
@@ -14,75 +27,57 @@ pub struct OrtEngine {
     session: Session,
     device: Device,
     inputs_minoptmax: Vec<Vec<MinOptMax>>,
-    inames: Vec<String>,
-    ishapes: Vec<Vec<isize>>,
-    idtypes: Vec<TensorElementType>,
-    onames: Vec<String>,
-    oshapes: Vec<Vec<isize>>,
-    odtypes: Vec<TensorElementType>,
+    inputs_attrs: OrtTensorAttr,
+    outputs_attrs: OrtTensorAttr,
     profile: bool,
     num_dry_run: usize,
+    model_proto: onnx::ModelProto,
+    params: usize,
+    wbmems: usize,
+    pub ts: Ts,
 }
 
 impl OrtEngine {
-    pub fn dry_run(&self) -> Result<()> {
-        if self.num_dry_run == 0 {
-            println!("{SAFE_CROSS_MARK} No dry run count specified, skipping the dry run.");
-            return Ok(());
-        }
-        let mut xs: Vec<Array<f32, IxDyn>> = Vec::new();
-        for i in self.inputs_minoptmax.iter() {
-            let mut x: Vec<usize> = Vec::new();
-            for i_ in i.iter() {
-                x.push(i_.opt as usize);
-            }
-            let x: Array<f32, IxDyn> = Array::ones(x).into_dyn();
-            xs.push(x);
-        }
-        for _ in 0..self.num_dry_run {
-            self.run(xs.as_ref())?;
-        }
-        println!("{CHECK_MARK} Dry run x{}", self.num_dry_run);
-        Ok(())
-    }
-
     pub fn new(config: &Options) -> Result<Self> {
-        ort::init().commit()?;
-        let session = Session::builder()?.with_model_from_file(&config.onnx_path)?;
+        // onnx graph
+        let model_proto = Self::load_onnx(&config.onnx_path)?;
+        let graph = match &model_proto.graph {
+            Some(graph) => graph,
+            None => anyhow::bail!("No graph found in this proto"),
+        };
 
-        // inputs
-        let mut ishapes = Vec::new();
-        let mut idtypes = Vec::new();
-        let mut inames = Vec::new();
-        for x in session.inputs.iter() {
-            inames.push(x.name.to_owned());
-            if let ValueType::Tensor { ty, dimensions } = &x.input_type {
-                ishapes.push(dimensions.iter().map(|x| *x as isize).collect::<Vec<_>>());
-                idtypes.push(*ty);
-            } else {
-                ishapes.push(vec![-1_isize]);
-                idtypes.push(ort::TensorElementType::Float32);
-            }
-        }
-        // outputs
-        let mut oshapes = Vec::new();
-        let mut odtypes = Vec::new();
-        let mut onames = Vec::new();
-        for x in session.outputs.iter() {
-            onames.push(x.name.to_owned());
-            if let ValueType::Tensor { ty, dimensions } = &x.output_type {
-                oshapes.push(dimensions.iter().map(|x| *x as isize).collect::<Vec<_>>());
-                odtypes.push(*ty);
-            } else {
-                oshapes.push(vec![-1_isize]);
-                odtypes.push(ort::TensorElementType::Float32);
-            }
+        // model params & mems
+        let byte_alignment = 16; // 16 for simd; 8 for most
+        let mut params: usize = 0;
+        let mut wbmems: usize = 0;
+        let mut initializer_names: HashSet<&str> = HashSet::new();
+        for tensor_proto in graph.initializer.iter() {
+            initializer_names.insert(&tensor_proto.name);
+            let param = tensor_proto.dims.iter().product::<i64>() as usize;
+            params += param;
+
+            // mems
+            let param = make_divisible(param, byte_alignment);
+            let n = Self::nbytes_from_onnx_dtype_id(tensor_proto.data_type as usize);
+            let wbmem = param * n;
+            wbmems += wbmem;
         }
+
+        // inputs & outputs
+        let inputs_attrs = Self::io_from_onnx_value_info(&initializer_names, &graph.input)?;
+        let outputs_attrs = Self::io_from_onnx_value_info(&initializer_names, &graph.output)?;
+
+        // inputs minoptmax
         let mut inputs_minoptmax: Vec<Vec<MinOptMax>> = Vec::new();
-        for (i, dims) in ishapes.iter().enumerate() {
+        for (i, dims) in inputs_attrs.dimss.iter().enumerate() {
             let mut v_: Vec<MinOptMax> = Vec::new();
             for (ii, &x) in dims.iter().enumerate() {
-                let x_default: MinOptMax = (ishapes[i][ii], ishapes[i][ii], ishapes[i][ii]).into();
+                let x_default: MinOptMax = (
+                    inputs_attrs.dimss[i][ii],
+                    inputs_attrs.dimss[i][ii],
+                    inputs_attrs.dimss[i][ii],
+                )
+                    .into();
                 let x: MinOptMax = match (i, ii) {
                     (0, 0) => Self::_set_ixx(x, &config.i00, i, ii).unwrap_or(x_default),
                     (0, 1) => Self::_set_ixx(x, &config.i01, i, ii).unwrap_or(x_default),
@@ -115,72 +110,80 @@ impl OrtEngine {
             inputs_minoptmax.push(v_);
         }
 
-        // build again
+        // build
+        ort::init().commit()?;
         let builder = Session::builder()?;
-        let device = config.device.to_owned();
-        let _ep = match device {
-            Device::Trt(device_id) => Self::build_trt(
-                &inames,
-                &inputs_minoptmax,
-                &builder,
-                device_id,
-                config.trt_int8_enable,
-                config.trt_fp16_enable,
-                config.trt_engine_cache_enable,
-            )?,
-            Device::Cuda(device_id) => Self::build_cuda(&builder, device_id)?,
-            Device::CoreML(_) => {
-                let coreml = ort::CoreMLExecutionProvider::default()
-                    .with_subgraphs()
-                    // .with_ane_only()
-                    .build();
-                if coreml.is_available()? && coreml.register(&builder).is_ok() {
-                    println!("{CHECK_MARK} Using CoreML");
-                    coreml
-                } else {
-                    println!("{CROSS_MARK} CoreML initialization failed");
-                    println!("{CHECK_MARK} Using CPU");
-                    ort::CPUExecutionProvider::default().build()
-                }
+        let mut device = config.device.to_owned();
+        match device {
+            Device::Trt(device_id) => {
+                Self::build_trt(
+                    &inputs_attrs.names,
+                    &inputs_minoptmax,
+                    &builder,
+                    device_id,
+                    config.trt_int8_enable,
+                    config.trt_fp16_enable,
+                    config.trt_engine_cache_enable,
+                )?;
             }
+            Device::Cuda(device_id) => {
+                Self::build_cuda(&builder, device_id).unwrap_or_else(|err| {
+                    device = Device::Cpu(0);
+                    println!("{err}");
+                })
+            }
+            Device::CoreML(_) => Self::build_coreml(&builder).unwrap_or_else(|err| {
+                device = Device::Cpu(0);
+                println!("{err}");
+            }),
             Device::Cpu(_) => {
-                println!("{CHECK_MARK} Using CPU");
-                ort::CPUExecutionProvider::default().build()
-            } // _ => todo!(),
-        };
+                Self::build_cpu(&builder)?;
+            }
+            _ => todo!(),
+        }
+
         let session = builder
             .with_optimization_level(ort::GraphOptimizationLevel::Level3)?
-            .with_model_from_file(&config.onnx_path)?;
+            .commit_from_file(&config.onnx_path)?;
+
+        // summary
+        println!(
+            "{CHECK_MARK} ORT: 1.{MINOR_VERSION}.x | Opset: {} | EP: {:?} | Dtype: {:?} | Parameters: {}",
+            model_proto.opset_import[0].version,
+            device,
+            inputs_attrs.dtypes,
+            human_bytes(params as f64),
+        );
 
         Ok(Self {
             session,
             device,
             inputs_minoptmax,
-            inames,
-            ishapes,
-            idtypes,
-            onames,
-            oshapes,
-            odtypes,
+            inputs_attrs,
+            outputs_attrs,
             profile: config.profile,
             num_dry_run: config.num_dry_run,
+            model_proto,
+            params,
+            wbmems,
+            ts: Ts::default(),
         })
     }
 
     fn build_trt(
-        inames: &[String],
+        names: &[String],
         inputs_minoptmax: &[Vec<MinOptMax>],
         builder: &SessionBuilder,
         device_id: usize,
         int8_enable: bool,
         fp16_enable: bool,
         engine_cache_enable: bool,
-    ) -> Result<ExecutionProviderDispatch> {
+    ) -> Result<()> {
         // auto generate shapes
         let mut spec_min = String::new();
         let mut spec_opt = String::new();
         let mut spec_max = String::new();
-        for (i, name) in inames.iter().enumerate() {
+        for (i, name) in names.iter().enumerate() {
             if i != 0 {
                 spec_min.push(',');
                 spec_opt.push(',');
@@ -217,81 +220,137 @@ impl OrtEngine {
             .with_timing_cache(false)
             .with_profile_min_shapes(spec_min)
             .with_profile_opt_shapes(spec_opt)
-            .with_profile_max_shapes(spec_max)
-            .build();
+            .with_profile_max_shapes(spec_max);
         if trt.is_available()? && trt.register(builder).is_ok() {
-            println!(
-                "{CHECK_MARK} Using TensorRT (Initial model serialization may require a wait)"
-            );
-            Ok(trt)
+            println!("\n🐢 Initial model serialization with TensorRT may require a wait...\n");
+            Ok(())
         } else {
-            println!("{CROSS_MARK} TensorRT initialization failed. Try CUDA...");
-            Self::build_cuda(builder, device_id)
+            anyhow::bail!("{CROSS_MARK} TensorRT initialization failed")
         }
     }
 
-    fn build_cuda(builder: &SessionBuilder, device_id: usize) -> Result<ExecutionProviderDispatch> {
-        let cuda = ort::CUDAExecutionProvider::default()
-            .with_device_id(device_id as i32)
-            .build();
-        if cuda.is_available()? && cuda.register(builder).is_ok() {
-            println!("{CHECK_MARK} Using CUDA");
-            Ok(cuda)
+    fn build_cuda(builder: &SessionBuilder, device_id: usize) -> Result<()> {
+        let ep = ort::CUDAExecutionProvider::default().with_device_id(device_id as i32);
+        if ep.is_available()? && ep.register(builder).is_ok() {
+            Ok(())
         } else {
-            println!("{CROSS_MARK} CUDA initialization failed");
-            println!("{CHECK_MARK} Using CPU");
-            Ok(ort::CPUExecutionProvider::default().build())
+            anyhow::bail!("{CROSS_MARK} CUDA initialization failed")
         }
     }
 
-    pub fn run(&self, xs: &[Array<f32, IxDyn>]) -> Result<Vec<Array<f32, IxDyn>>> {
-        // input
+    fn build_coreml(builder: &SessionBuilder) -> Result<()> {
+        let ep = ort::CoreMLExecutionProvider::default().with_subgraphs(); //.with_ane_only();
+        if ep.is_available()? && ep.register(builder).is_ok() {
+            Ok(())
+        } else {
+            anyhow::bail!("{CROSS_MARK} CoreML initialization failed")
+        }
+    }
+
+    fn build_cpu(builder: &SessionBuilder) -> Result<()> {
+        let ep = ort::CUDAExecutionProvider::default();
+        if ep.is_available()? && ep.register(builder).is_ok() {
+            Ok(())
+        } else {
+            anyhow::bail!("{CROSS_MARK} CPU initialization failed")
+        }
+    }
+
+    pub fn dry_run(&mut self) -> Result<()> {
+        if self.num_dry_run > 0 {
+            let mut xs: Vec<Array<f32, IxDyn>> = Vec::new();
+            for i in self.inputs_minoptmax.iter() {
+                let mut x: Vec<usize> = Vec::new();
+                for i_ in i.iter() {
+                    x.push(i_.opt as usize);
+                }
+                let x: Array<f32, IxDyn> = Array::ones(x).into_dyn();
+                xs.push(x);
+            }
+            for _ in 0..self.num_dry_run {
+                self.run(xs.as_ref())?;
+            }
+            self.ts.clear();
+            println!("{CHECK_MARK} Dryrun x{}", self.num_dry_run);
+        }
+        Ok(())
+    }
+
+    pub fn run(&mut self, xs: &[Array<f32, IxDyn>]) -> Result<Vec<Array<f32, IxDyn>>> {
+        // inputs dtype alignment
         let mut xs_ = Vec::new();
         let t_pre = std::time::Instant::now();
-        for (idtype, x) in self.idtypes.iter().zip(xs.iter()) {
-            let x_ = match idtype {
-                TensorElementType::Float32 => ort::Value::from_array(x.view())?,
-                TensorElementType::Float16 => ort::Value::from_array(x.mapv(f16::from_f32).view())?,
-                TensorElementType::Int32 => ort::Value::from_array(x.mapv(|x_| x_ as i32).view())?,
-                TensorElementType::Int64 => ort::Value::from_array(x.mapv(|x_| x_ as i64).view())?,
+        for (idtype, x) in self.inputs_attrs.dtypes.iter().zip(xs.iter()) {
+            let x_ = match &idtype {
+                TensorElementType::Float32 => ort::Value::from_array(x.view())?.into_dyn(),
+                TensorElementType::Float16 => {
+                    ort::Value::from_array(x.mapv(f16::from_f32).view())?.into_dyn()
+                }
+                TensorElementType::Int32 => {
+                    ort::Value::from_array(x.mapv(|x_| x_ as i32).view())?.into_dyn()
+                }
+                TensorElementType::Int64 => {
+                    ort::Value::from_array(x.mapv(|x_| x_ as i64).view())?.into_dyn()
+                }
                 _ => todo!(),
             };
-            xs_.push(x_);
+            xs_.push(Into::<ort::SessionInputValue<'_>>::into(x_));
         }
         let t_pre = t_pre.elapsed();
+        self.ts.add_or_push(0, t_pre);
 
         // inference
         let t_run = std::time::Instant::now();
-        let ys = self.session.run(xs_.as_ref())?;
+        let outputs = self.session.run(&xs_[..])?;
         let t_run = t_run.elapsed();
+        self.ts.add_or_push(1, t_run);
 
         // oputput
-        let mut ys_ = Vec::new();
+        let mut ys = Vec::new();
         let t_post = std::time::Instant::now();
-
-        for (dtype, name) in self.odtypes.iter().zip(self.onames.iter()) {
-            let y = &ys[name.as_str()];
+        for (dtype, name) in self
+            .outputs_attrs
+            .dtypes
+            .iter()
+            .zip(self.outputs_attrs.names.iter())
+        {
+            let y = &outputs[name.as_str()];
             let y_ = match &dtype {
-                TensorElementType::Float32 => y.extract_tensor::<f32>()?.view().to_owned(),
-                TensorElementType::Float16 => y.extract_tensor::<f16>()?.view().mapv(f16::to_f32),
+                TensorElementType::Float32 => y.try_extract_tensor::<f32>()?.view().into_owned(),
+                TensorElementType::Float16 => y
+                    .try_extract_tensor::<f16>()?
+                    .view()
+                    .mapv(f16::to_f32)
+                    .into_owned(),
                 TensorElementType::Int64 => y
-                    .extract_tensor::<i64>()?
+                    .try_extract_tensor::<i64>()?
                     .view()
                     .to_owned()
-                    .mapv(|x| x as f32),
+                    .mapv(|x| x as f32)
+                    .into_owned(),
                 _ => todo!(),
             };
-            ys_.push(y_);
+            ys.push(y_);
         }
         let t_post = t_post.elapsed();
+        self.ts.add_or_push(2, t_post);
+
         if self.profile {
+            let len = 10usize;
+            let n = 4usize;
             println!(
-                "[Profile] batch: {:?} => {:.4?} (i: {t_pre:.4?}, run: {t_run:.4?}, o: {t_post:.4?})", 
-                self.batch().opt,
-                t_pre + t_run + t_post
+                "[Profile] {:>len$.n$?} ({:>len$.n$?} avg) [alignment: {:>len$.n$?} ({:>len$.n$?} avg) | inference: {:>len$.n$?} ({:>len$.n$?} avg) | to_f32: {:>len$.n$?} ({:>len$.n$?} avg)]",
+                t_pre + t_run + t_post,
+                self.ts.avg(),
+                t_pre,
+                self.ts.avgi(0),
+                t_run,
+                self.ts.avgi(1),
+                t_post,
+                self.ts.avgi(2),
             );
         }
-        Ok(ys_)
+        Ok(ys)
     }
 
     pub fn _set_ixx(x: isize, ixx: &Option<MinOptMax>, i: usize, ii: usize) -> Option<MinOptMax> {
@@ -310,28 +369,200 @@ impl OrtEngine {
         }
     }
 
+    pub fn nbytes_from_onnx_dtype_id(x: usize) -> usize {
+        match x {
+            7 | 11 | 13 => 8,     // i64, f64, u64
+            1 | 6 | 12 => 4,      // f32, i32, u32
+            10 | 16 | 5 | 4 => 2, // f16, bf16, i16, u16
+            2 | 3 | 9 => 1,       // u8, i8, bool
+            8 => 4,               // string(1~4)
+            _ => todo!(),
+        }
+    }
+
+    pub fn nbytes_from_onnx_dtype(x: &ort::TensorElementType) -> usize {
+        match x {
+            ort::TensorElementType::Float64
+            | ort::TensorElementType::Uint64
+            | ort::TensorElementType::Int64 => 8, // i64, f64, u64
+            ort::TensorElementType::Float32
+            | ort::TensorElementType::Uint32
+            | ort::TensorElementType::Int32
+            | ort::TensorElementType::String => 4, // f32, i32, u32, string(1~4)
+            ort::TensorElementType::Float16
+            | ort::TensorElementType::Bfloat16
+            | ort::TensorElementType::Int16
+            | ort::TensorElementType::Uint16 => 2, // f16, bf16, i16, u16
+            ort::TensorElementType::Uint8
+            | ort::TensorElementType::Int8
+            | ort::TensorElementType::Bool => 1, // u8, i8, bool
+        }
+    }
+
+    fn ort_dtype_from_onnx_dtype_id(value: i32) -> Option<ort::TensorElementType> {
+        match value {
+            0 => None,
+            1 => Some(ort::TensorElementType::Float32),
+            2 => Some(ort::TensorElementType::Uint8),
+            3 => Some(ort::TensorElementType::Int8),
+            4 => Some(ort::TensorElementType::Uint16),
+            5 => Some(ort::TensorElementType::Int16),
+            6 => Some(ort::TensorElementType::Int32),
+            7 => Some(ort::TensorElementType::Int64),
+            8 => Some(ort::TensorElementType::String),
+            9 => Some(ort::TensorElementType::Bool),
+            10 => Some(ort::TensorElementType::Float16),
+            11 => Some(ort::TensorElementType::Float64),
+            12 => Some(ort::TensorElementType::Uint32),
+            13 => Some(ort::TensorElementType::Uint64),
+            14 => None, // COMPLEX64
+            15 => None, // COMPLEX128
+            16 => Some(ort::TensorElementType::Bfloat16),
+            _ => None,
+        }
+    }
+
+    #[allow(dead_code)]
+    fn i_from_session(session: &ort::Session) -> Result<OrtTensorAttr> {
+        let mut dimss = Vec::new();
+        let mut dtypes = Vec::new();
+        let mut names = Vec::new();
+        for x in session.inputs.iter() {
+            names.push(x.name.to_owned());
+            if let ort::ValueType::Tensor { ty, dimensions } = &x.input_type {
+                dimss.push(dimensions.iter().map(|x| *x as isize).collect::<Vec<_>>());
+                dtypes.push(*ty);
+            } else {
+                dimss.push(vec![-1_isize]);
+                dtypes.push(ort::TensorElementType::Float32);
+            }
+        }
+
+        Ok(OrtTensorAttr {
+            names,
+            dimss,
+            dtypes,
+        })
+    }
+
+    #[allow(dead_code)]
+    fn o_from_session(session: &ort::Session) -> Result<OrtTensorAttr> {
+        let mut dimss = Vec::new();
+        let mut dtypes = Vec::new();
+        let mut names = Vec::new();
+        for x in session.outputs.iter() {
+            names.push(x.name.to_owned());
+            if let ort::ValueType::Tensor { ty, dimensions } = &x.output_type {
+                dimss.push(dimensions.iter().map(|x| *x as isize).collect::<Vec<_>>());
+                dtypes.push(*ty);
+            } else {
+                dimss.push(vec![-1_isize]);
+                dtypes.push(ort::TensorElementType::Float32);
+            }
+        }
+
+        Ok(OrtTensorAttr {
+            names,
+            dimss,
+            dtypes,
+        })
+    }
+
+    fn io_from_onnx_value_info(
+        initializer_names: &HashSet<&str>,
+        value_info: &[onnx::ValueInfoProto],
+    ) -> Result<OrtTensorAttr> {
+        let mut dimss: Vec<Vec<isize>> = Vec::new();
+        let mut dtypes: Vec<ort::TensorElementType> = Vec::new();
+        let mut names: Vec<String> = Vec::new();
+        for v in value_info.iter() {
+            if initializer_names.contains(v.name.as_str()) {
+                continue;
+            }
+            names.push(v.name.to_string());
+            let dtype = match &v.r#type {
+                Some(dtype) => dtype,
+                None => continue,
+            };
+            let dtype = match &dtype.value {
+                Some(dtype) => dtype,
+                None => continue,
+            };
+            let tensor = match dtype {
+                onnx::type_proto::Value::TensorType(tensor) => tensor,
+                _ => continue,
+            };
+            let tensor_type = tensor.elem_type;
+            let tensor_type = match Self::ort_dtype_from_onnx_dtype_id(tensor_type) {
+                Some(dtype) => dtype,
+                None => continue,
+                // None => anyhow::bail!("DType not supported"),
+            };
+            dtypes.push(tensor_type);
+
+            let shapes = match &tensor.shape {
+                Some(shapes) => shapes,
+                None => continue,
+                // None => anyhow::bail!("DType has no shapes"),
+            };
+            let mut shape_: Vec<isize> = Vec::new();
+            for shape in shapes.dim.iter() {
+                match &shape.value {
+                    None => continue,
+                    Some(value) => match value {
+                        onnx::tensor_shape_proto::dimension::Value::DimValue(x) => {
+                            shape_.push(*x as isize);
+                        }
+                        onnx::tensor_shape_proto::dimension::Value::DimParam(_) => {
+                            shape_.push(-1isize);
+                        }
+                    },
+                }
+            }
+            dimss.push(shape_);
+        }
+        Ok(OrtTensorAttr {
+            dimss,
+            dtypes,
+            names,
+        })
+    }
+
+    pub fn load_onnx<P: AsRef<std::path::Path>>(p: P) -> Result<onnx::ModelProto> {
+        let f = std::fs::read(p)?;
+        Ok(onnx::ModelProto::decode(f.as_slice())?)
+    }
+
     pub fn oshapes(&self) -> &Vec<Vec<isize>> {
-        &self.oshapes
+        &self.outputs_attrs.dimss
+    }
+
+    pub fn odimss(&self) -> &Vec<Vec<isize>> {
+        &self.outputs_attrs.dimss
     }
 
     pub fn onames(&self) -> &Vec<String> {
-        &self.onames
+        &self.outputs_attrs.names
     }
 
     pub fn odtypes(&self) -> &Vec<ort::TensorElementType> {
-        &self.odtypes
+        &self.outputs_attrs.dtypes
     }
 
     pub fn ishapes(&self) -> &Vec<Vec<isize>> {
-        &self.ishapes
+        &self.inputs_attrs.dimss
+    }
+
+    pub fn idimss(&self) -> &Vec<Vec<isize>> {
+        &self.inputs_attrs.dimss
     }
 
     pub fn inames(&self) -> &Vec<String> {
-        &self.inames
+        &self.inputs_attrs.names
     }
 
     pub fn idtypes(&self) -> &Vec<ort::TensorElementType> {
-        &self.idtypes
+        &self.inputs_attrs.dtypes
     }
 
     pub fn device(&self) -> &Device {
@@ -355,7 +586,7 @@ impl OrtEngine {
     }
 
     pub fn is_batch_dyn(&self) -> bool {
-        self.ishapes[0][0] == -1
+        self.ishapes()[0][0] == -1
     }
 
     pub fn try_fetch(&self, key: &str) -> Option<String> {
@@ -372,7 +603,31 @@ impl OrtEngine {
         &self.session
     }
 
-    pub fn version(&self) -> Option<String> {
-        self.try_fetch("version")
+    pub fn ir_version(&self) -> usize {
+        self.model_proto.ir_version as usize
+    }
+
+    pub fn opset_version(&self) -> usize {
+        self.model_proto.opset_import[0].version as usize
+    }
+
+    pub fn producer_name(&self) -> String {
+        self.model_proto.producer_name.to_string()
+    }
+
+    pub fn producer_version(&self) -> String {
+        self.model_proto.producer_version.to_string()
+    }
+
+    pub fn model_version(&self) -> usize {
+        self.model_proto.model_version as usize
+    }
+
+    pub fn parameters(&self) -> usize {
+        self.params
+    }
+
+    pub fn memory_weights(&self) -> usize {
+        self.wbmems
     }
 }

src/core/min_opt_max.rs

@@ -1,5 +1,5 @@
 /// A value composed of Min-Opt-Max
-#[derive(Debug, Clone)]
+#[derive(Clone)]
 pub struct MinOptMax {
     pub min: isize,
     pub opt: isize,
@@ -16,6 +16,16 @@ impl Default for MinOptMax {
     }
 }
 
+impl std::fmt::Debug for MinOptMax {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("")
+            .field("Min", &self.min)
+            .field("Opt", &self.opt)
+            .field("Max", &self.max)
+            .finish()
+    }
+}
+
 impl From<(isize, isize, isize)> for MinOptMax {
     fn from((min, opt, max): (isize, isize, isize)) -> Self {
         let min = min.min(opt);

src/core/mod.rs

@@ -6,9 +6,11 @@ mod engine;
 mod logits_sampler;
 mod metric;
 mod min_opt_max;
+pub mod onnx;
 pub mod ops;
 mod options;
 mod tokenizer_stream;
+mod ts;
 
 pub use annotator::Annotator;
 pub use dataloader::DataLoader;
@@ -20,3 +22,4 @@ pub use metric::Metric;
 pub use min_opt_max::MinOptMax;
 pub use options::Options;
 pub use tokenizer_stream::TokenizerStream;
+pub use ts::Ts;

src/core/ops.rs

@@ -1,6 +1,7 @@
 use anyhow::Result;
+use fast_image_resize as fr;
 use image::{DynamicImage, GenericImageView, ImageBuffer};
-use ndarray::{Array, Axis, IxDyn};
+use ndarray::{s, Array, Axis, IxDyn};
 
 pub fn standardize(xs: Array<f32, IxDyn>, mean: &[f32], std: &[f32]) -> Array<f32, IxDyn> {
     let mean = Array::from_shape_vec((1, mean.len(), 1, 1), mean.to_vec()).unwrap();
@@ -26,18 +27,57 @@ pub fn scale_wh(w0: f32, h0: f32, w1: f32, h1: f32) -> (f32, f32, f32) {
     (r, (w0 * r).round(), (h0 * r).round())
 }
 
-pub fn resize(xs: &[DynamicImage], height: u32, width: u32) -> Result<Array<f32, IxDyn>> {
+pub fn build_resizer(ty: &str) -> fr::Resizer {
+    let ty = match ty {
+        "box" => fr::FilterType::Box,
+        "bilinear" => fr::FilterType::Bilinear,
+        "hamming" => fr::FilterType::Hamming,
+        "catmullRom" => fr::FilterType::CatmullRom,
+        "mitchell" => fr::FilterType::Mitchell,
+        "lanczos3" => fr::FilterType::Lanczos3,
+        _ => todo!(),
+    };
+    fr::Resizer::new(fr::ResizeAlg::Convolution(ty))
+}
+
+pub fn resize(
+    xs: &[DynamicImage],
+    height: u32,
+    width: u32,
+    filter: &str,
+) -> Result<Array<f32, IxDyn>> {
     let mut ys = Array::ones((xs.len(), 3, height as usize, width as usize)).into_dyn();
+    let mut resizer = build_resizer(filter);
     for (idx, x) in xs.iter().enumerate() {
-        let img = x.resize_exact(width, height, image::imageops::FilterType::Triangle);
-        for (x, y, rgb) in img.pixels() {
-            let x = x as usize;
-            let y = y as usize;
-            let [r, g, b, _] = rgb.0;
-            ys[[idx, 0, y, x]] = r as f32;
-            ys[[idx, 1, y, x]] = g as f32;
-            ys[[idx, 2, y, x]] = b as f32;
-        }
+        // src
+        let src_image = fr::Image::from_vec_u8(
+            std::num::NonZeroU32::new(x.width()).unwrap(),
+            std::num::NonZeroU32::new(x.height()).unwrap(),
+            x.to_rgb8().into_raw(),
+            fr::PixelType::U8x3,
+        )
+        .unwrap();
+
+        // dst
+        let mut dst_image = fr::Image::new(
+            std::num::NonZeroU32::new(width).unwrap(),
+            std::num::NonZeroU32::new(height).unwrap(),
+            src_image.pixel_type(),
+        );
+
+        // resize
+        resizer
+            .resize(&src_image.view(), &mut dst_image.view_mut())
+            .unwrap();
+        let buffer = dst_image.into_vec();
+
+        // to ndarray
+        let y_ = Array::from_shape_vec((height as usize, width as usize, 3), buffer)
+            .unwrap()
+            .mapv(|x| x as f32)
+            .permuted_axes([2, 0, 1]);
+        let mut data = ys.slice_mut(s![idx, .., .., ..]);
+        data.assign(&y_);
     }
     Ok(ys)
 }
@@ -46,27 +86,62 @@ pub fn letterbox(
     xs: &[DynamicImage],
     height: u32,
     width: u32,
-    bg: f32,
+    filter: &str,
+    bg: Option<u8>,
 ) -> Result<Array<f32, IxDyn>> {
-    // TODO: refactor
     let mut ys = Array::ones((xs.len(), 3, height as usize, width as usize)).into_dyn();
-    ys.fill(bg);
+    let mut resizer = build_resizer(filter);
     for (idx, x) in xs.iter().enumerate() {
         let (w0, h0) = x.dimensions();
         let (_, w_new, h_new) = scale_wh(w0 as f32, h0 as f32, width as f32, height as f32);
-        let img = x.resize_exact(
-            w_new as u32,
-            h_new as u32,
-            image::imageops::FilterType::CatmullRom,
-        );
-        for (x, y, rgb) in img.pixels() {
-            let x = x as usize;
-            let y = y as usize;
-            let [r, g, b, _] = rgb.0;
-            ys[[idx, 0, y, x]] = r as f32;
-            ys[[idx, 1, y, x]] = g as f32;
-            ys[[idx, 2, y, x]] = b as f32;
-        }
+
+        // src
+        let src_image = fr::Image::from_vec_u8(
+            std::num::NonZeroU32::new(w0).unwrap(),
+            std::num::NonZeroU32::new(h0).unwrap(),
+            x.to_rgb8().into_raw(),
+            fr::PixelType::U8x3,
+        )
+        .unwrap();
+
+        // dst
+        let mut dst_image = match bg {
+            Some(bg) => fr::Image::from_vec_u8(
+                std::num::NonZeroU32::new(width).unwrap(),
+                std::num::NonZeroU32::new(height).unwrap(),
+                vec![bg; 3 * height as usize * width as usize],
+                src_image.pixel_type(),
+            )
+            .unwrap(),
+            None => fr::Image::new(
+                std::num::NonZeroU32::new(width).unwrap(),
+                std::num::NonZeroU32::new(height).unwrap(),
+                src_image.pixel_type(),
+            ),
+        };
+
+        // mutable view
+        let mut dst_view = dst_image
+            .view_mut()
+            .crop(
+                0,
+                0,
+                std::num::NonZeroU32::new(w_new as u32).unwrap(),
+                std::num::NonZeroU32::new(h_new as u32).unwrap(),
+            )
+            .unwrap();
+
+        // resize
+        resizer.resize(&src_image.view(), &mut dst_view).unwrap();
+        let buffer = dst_image.into_vec();
+
+        // to ndarray
+        let y_ = Array::from_shape_vec((height as usize, width as usize, 3), buffer)
+            .unwrap()
+            .mapv(|x| x as f32)
+            .permuted_axes([2, 0, 1]);
+        let mut data = ys.slice_mut(s![idx, .., .., ..]);
+        data.assign(&y_);
     }
     Ok(ys)
 }
@@ -75,23 +150,63 @@ pub fn resize_with_fixed_height(
     xs: &[DynamicImage],
     height: u32,
     width: u32,
-    bg: f32,
+    filter: &str,
+    bg: Option<u8>,
 ) -> Result<Array<f32, IxDyn>> {
     let mut ys = Array::ones((xs.len(), 3, height as usize, width as usize)).into_dyn();
-    ys.fill(bg);
+    let mut resizer = build_resizer(filter);
     for (idx, x) in xs.iter().enumerate() {
         let (w0, h0) = x.dimensions();
         let h_new = height;
         let w_new = height * w0 / h0;
-        let img = x.resize_exact(w_new, h_new, image::imageops::FilterType::CatmullRom);
-        for (x, y, rgb) in img.pixels() {
-            let x = x as usize;
-            let y = y as usize;
-            let [r, g, b, _] = rgb.0;
-            ys[[idx, 0, y, x]] = r as f32;
-            ys[[idx, 1, y, x]] = g as f32;
-            ys[[idx, 2, y, x]] = b as f32;
-        }
+
+        // src
+        let src_image = fr::Image::from_vec_u8(
+            std::num::NonZeroU32::new(w0).unwrap(),
+            std::num::NonZeroU32::new(h0).unwrap(),
+            x.to_rgb8().into_raw(),
+            fr::PixelType::U8x3,
+        )
+        .unwrap();
+
+        // dst
+        let mut dst_image = match bg {
+            Some(bg) => fr::Image::from_vec_u8(
+                std::num::NonZeroU32::new(width).unwrap(),
+                std::num::NonZeroU32::new(height).unwrap(),
+                vec![bg; 3 * height as usize * width as usize],
+                src_image.pixel_type(),
+            )
+            .unwrap(),
+            None => fr::Image::new(
+                std::num::NonZeroU32::new(width).unwrap(),
+                std::num::NonZeroU32::new(height).unwrap(),
+                src_image.pixel_type(),
+            ),
+        };
+
+        // mutable view
+        let mut dst_view = dst_image
+            .view_mut()
+            .crop(
+                0,
+                0,
+                std::num::NonZeroU32::new(w_new).unwrap(),
+                std::num::NonZeroU32::new(h_new).unwrap(),
+            )
+            .unwrap();
+
+        // resize
+        resizer.resize(&src_image.view(), &mut dst_view).unwrap();
+        let buffer = dst_image.into_vec();
+
+        // to ndarray
+        let y_ = Array::from_shape_vec((height as usize, width as usize, 3), buffer)
+            .unwrap()
+            .mapv(|x| x as f32)
+            .permuted_axes([2, 0, 1]);
+        let mut data = ys.slice_mut(s![idx, .., .., ..]);
+        data.assign(&y_);
     }
     Ok(ys)
 }
@@ -109,3 +224,7 @@ pub fn descale_mask(mask: DynamicImage, w0: f32, h0: f32, w1: f32, h1: f32) -> D
     let mask = mask.crop(0, 0, w as u32, h as u32);
     mask.resize_exact(w1 as u32, h1 as u32, image::imageops::FilterType::Triangle)
 }
+
+pub fn make_divisible(x: usize, divisor: usize) -> usize {
+    (x - 1 + divisor) / divisor * divisor
+}

src/core/options.rs

@@ -67,7 +67,7 @@ impl Default for Options {
             onnx_path: String::new(),
             device: Device::Cuda(0),
             profile: false,
-            num_dry_run: 3,
+            num_dry_run: 5,
             i00: None,
             i01: None,
             i02: None,

src/core/ts.rs

@@ -0,0 +1,55 @@
+use std::time::Duration;
+
+#[derive(Debug, Default)]
+pub struct Ts {
+    n: usize,
+    ts: Vec<Duration>,
+    // pub t0: Duration,
+    // pub t1: Duration,
+    // pub t2: Duration,
+    // pub t3: Duration,
+    // pub t4: Duration,
+}
+
+impl Ts {
+    pub fn total(&self) -> Duration {
+        self.ts.iter().sum::<Duration>()
+    }
+
+    pub fn n(&self) -> usize {
+        self.n / self.ts.len()
+    }
+
+    pub fn avg(&self) -> Duration {
+        self.total() / self.n() as u32
+    }
+
+    pub fn avgi(&self, i: usize) -> Duration {
+        if i >= self.ts.len() {
+            panic!("Index out of bound");
+        }
+        self.ts[i] / self.n() as u32
+    }
+
+    pub fn ts(&self) -> &Vec<Duration> {
+        &self.ts
+    }
+
+    // TODO: overhead?
+    pub fn add_or_push(&mut self, i: usize, x: Duration) {
+        match self.ts.get_mut(i) {
+            Some(elem) => *elem += x,
+            None => {
+                if i >= self.ts.len() {
+                    self.ts.push(x)
+                }
+            }
+        }
+        self.n += 1;
+    }
+
+    pub fn clear(&mut self) {
+        self.n = Default::default();
+        self.ts = Default::default();
+    }
+}

src/models/blip.rs

@@ -19,8 +19,8 @@ pub struct Blip {
 
 impl Blip {
     pub fn new(options_visual: Options, options_textual: Options) -> Result<Self> {
-        let visual = OrtEngine::new(&options_visual)?;
-        let textual = OrtEngine::new(&options_textual)?;
+        let mut visual = OrtEngine::new(&options_visual)?;
+        let mut textual = OrtEngine::new(&options_textual)?;
         let (batch_visual, batch_textual, height, width) = (
             visual.batch().to_owned(),
             textual.batch().to_owned(),
@@ -42,18 +42,21 @@ impl Blip {
         })
     }
 
-    pub fn encode_images(&self, xs: &[DynamicImage]) -> Result<Embedding> {
-        let xs_ = ops::resize(xs, self.height.opt as u32, self.width.opt as u32)?;
-        let xs_ = ops::normalize(xs_, 0.0, 255.0);
+    pub fn encode_images(&mut self, xs: &[DynamicImage]) -> Result<Embedding> {
+        let xs_ = ops::resize(
+            xs,
+            self.height.opt as u32,
+            self.width.opt as u32,
+            "bilinear",
+        )?;
+        let xs_ = ops::normalize(xs_, 0., 255.);
         let xs_ = ops::standardize(
             xs_,
             &[0.48145466, 0.4578275, 0.40821073],
             &[0.26862954, 0.2613026, 0.2757771],
         );
         let ys: Vec<Array<f32, IxDyn>> = self.visual.run(&[xs_])?;
-        // let ys = ys[0].to_owned();
         Ok(Embedding::new(ys[0].to_owned()))
-        // Ok(ys)
     }
 
     pub fn caption(

src/models/clip.rs

@@ -19,8 +19,8 @@ pub struct Clip {
 impl Clip {
     pub fn new(options_visual: Options, options_textual: Options) -> Result<Self> {
         let context_length = 77;
-        let visual = OrtEngine::new(&options_visual)?;
-        let textual = OrtEngine::new(&options_textual)?;
+        let mut visual = OrtEngine::new(&options_visual)?;
+        let mut textual = OrtEngine::new(&options_textual)?;
         let (batch_visual, batch_textual, height, width) = (
             visual.inputs_minoptmax()[0][0].to_owned(),
             textual.inputs_minoptmax()[0][0].to_owned(),
@@ -52,9 +52,14 @@ impl Clip {
         })
     }
 
-    pub fn encode_images(&self, xs: &[DynamicImage]) -> Result<Embedding> {
-        let xs_ = ops::resize(xs, self.height.opt as u32, self.width.opt as u32)?;
-        let xs_ = ops::normalize(xs_, 0.0, 255.0);
+    pub fn encode_images(&mut self, xs: &[DynamicImage]) -> Result<Embedding> {
+        let xs_ = ops::resize(
+            xs,
+            self.height.opt as u32,
+            self.width.opt as u32,
+            "bilinear",
+        )?;
+        let xs_ = ops::normalize(xs_, 0., 255.);
         let xs_ = ops::standardize(
             xs_,
             &[0.48145466, 0.4578275, 0.40821073],
@@ -64,7 +69,7 @@ impl Clip {
         Ok(Embedding::new(ys[0].to_owned()))
     }
 
-    pub fn encode_texts(&self, texts: &[String]) -> Result<Embedding> {
+    pub fn encode_texts(&mut self, texts: &[String]) -> Result<Embedding> {
         let encodings = self
             .tokenizer
             .encode_batch(texts.to_owned(), false)

src/models/db.rs

@@ -17,8 +17,8 @@ pub struct DB {
 }
 
 impl DB {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.batch().to_owned(),
             engine.height().to_owned(),
@@ -27,8 +27,8 @@ impl DB {
         let confs = DynConf::new(&options.confs, 1);
         let unclip_ratio = options.unclip_ratio;
         let binary_thresh = 0.2;
-        let min_width = options.min_width.unwrap_or(0.0);
-        let min_height = options.min_height.unwrap_or(0.0);
+        let min_width = options.min_width.unwrap_or(0.);
+        let min_height = options.min_height.unwrap_or(0.);
         engine.dry_run()?;
 
         Ok(Self {
@@ -45,8 +45,14 @@ impl DB {
     }
 
     pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
-        let xs_ = ops::letterbox(xs, self.height.opt as u32, self.width.opt as u32, 144.0)?;
-        let xs_ = ops::normalize(xs_, 0.0, 255.0);
+        let xs_ = ops::letterbox(
+            xs,
+            self.height.opt as u32,
+            self.width.opt as u32,
+            "bilinear",
+            Some(114),
+        )?;
+        let xs_ = ops::normalize(xs_, 0., 255.);
         let xs_ = ops::standardize(xs_, &[0.485, 0.456, 0.406], &[0.229, 0.224, 0.225]);
         let ys = self.engine.run(&[xs_])?;
         self.postprocess(ys, xs)

src/models/depth_anything.rs

@@ -12,8 +12,8 @@ pub struct DepthAnything {
 }
 
 impl DepthAnything {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.batch().to_owned(),
             engine.height().to_owned(),
@@ -29,8 +29,13 @@ impl DepthAnything {
         })
     }
 
-    pub fn run(&self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
-        let xs_ = ops::resize(xs, self.height.opt as u32, self.width.opt as u32)?;
+    pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
+        let xs_ = ops::resize(
+            xs,
+            self.height.opt as u32,
+            self.width.opt as u32,
+            "lanczos3",
+        )?;
         let xs_ = ops::normalize(xs_, 0.0, 255.0);
         let xs_ = ops::standardize(xs_, &[0.485, 0.456, 0.406], &[0.229, 0.224, 0.225]);
         let ys = self.engine.run(&[xs_])?;

src/models/dinov2.rs

@@ -21,14 +21,14 @@ pub struct Dinov2 {
 }
 
 impl Dinov2 {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.inputs_minoptmax()[0][0].to_owned(),
             engine.inputs_minoptmax()[0][2].to_owned(),
             engine.inputs_minoptmax()[0][3].to_owned(),
         );
-        let which = match &options.onnx_path {
+        let which = match options.onnx_path {
             s if s.contains("b14") => Model::B,
             s if s.contains("s14") => Model::S,
             _ => todo!(),
@@ -49,7 +49,12 @@ impl Dinov2 {
     }
 
     pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Array<f32, IxDyn>> {
-        let xs_ = ops::resize(xs, self.height.opt as u32, self.width.opt as u32)?;
+        let xs_ = ops::resize(
+            xs,
+            self.height.opt as u32,
+            self.width.opt as u32,
+            "lanczos3",
+        )?;
         let xs_ = ops::normalize(xs_, 0.0, 255.0);
         let xs_ = ops::standardize(
             xs_,

src/models/modnet.rs

@@ -13,8 +13,8 @@ pub struct MODNet {
 }
 
 impl MODNet {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.batch().to_owned(),
             engine.height().to_owned(),
@@ -30,9 +30,14 @@ impl MODNet {
         })
     }
 
-    pub fn run(&self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
-        let xs_ = ops::resize(xs, self.height.opt as u32, self.width.opt as u32)?;
-        let xs_ = ops::normalize(xs_, 127.5, 255.0);
+    pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
+        let xs_ = ops::resize(
+            xs,
+            self.height.opt as u32,
+            self.width.opt as u32,
+            "lanczos3",
+        )?;
+        let xs_ = ops::normalize(xs_, 127.5, 255.);
         let ys = self.engine.run(&[xs_])?;
         self.postprocess(ys, xs)
     }

src/models/rtdetr.rs

@@ -17,14 +17,14 @@ pub struct RTDETR {
 }
 
 impl RTDETR {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.inputs_minoptmax()[0][0].to_owned(),
             engine.inputs_minoptmax()[0][2].to_owned(),
             engine.inputs_minoptmax()[0][3].to_owned(),
         );
-        let names: Option<_> = match &options.names {
+        let names: Option<_> = match options.names {
             None => engine.try_fetch("names").map(|names| {
                 let re = Regex::new(r#"(['"])([-()\w '"]+)(['"])"#).unwrap();
                 let mut names_ = vec![];
@@ -56,7 +56,13 @@ impl RTDETR {
     }
 
     pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
-        let xs_ = ops::letterbox(xs, self.height() as u32, self.width() as u32, 144.0)?;
+        let xs_ = ops::letterbox(
+            xs,
+            self.height() as u32,
+            self.width() as u32,
+            "catmullRom",
+            Some(114),
+        )?;
         let xs_ = ops::normalize(xs_, 0.0, 255.0);
         let ys = self.engine.run(&[xs_])?;
         self.postprocess(ys, xs)

src/models/rtmo.rs

@@ -15,8 +15,8 @@ pub struct RTMO {
 }
 
 impl RTMO {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.batch().to_owned(),
             engine.height().to_owned(),
@@ -39,7 +39,13 @@ impl RTMO {
     }
 
     pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
-        let xs_ = ops::letterbox(xs, self.height() as u32, self.width() as u32, 114.0)?;
+        let xs_ = ops::letterbox(
+            xs,
+            self.height() as u32,
+            self.width() as u32,
+            "catmullRom",
+            Some(114),
+        )?;
         let ys = self.engine.run(&[xs_])?;
         self.postprocess(ys, xs)
     }

src/models/svtr.rs

@@ -15,8 +15,8 @@ pub struct SVTR {
 }
 
 impl SVTR {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.batch().to_owned(),
             engine.height().to_owned(),
@@ -24,7 +24,7 @@ impl SVTR {
         );
         let confs = DynConf::new(&options.confs, 1);
         let mut vocab: Vec<_> =
-            std::fs::read_to_string(options.vocab.as_ref().expect("No vocabulary found"))?
+            std::fs::read_to_string(options.vocab.expect("No vocabulary found"))?
                 .lines()
                 .map(|line| line.to_string())
                 .collect();
@@ -43,8 +43,13 @@ impl SVTR {
     }
 
     pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
-        let xs_ =
-            ops::resize_with_fixed_height(xs, self.height.opt as u32, self.width.opt as u32, 0.0)?;
+        let xs_ = ops::resize_with_fixed_height(
+            xs,
+            self.height.opt as u32,
+            self.width.opt as u32,
+            "bilinear",
+            Some(0),
+        )?;
         let xs_ = ops::normalize(xs_, 0.0, 255.0);
         let ys: Vec<Array<f32, IxDyn>> = self.engine.run(&[xs_])?;
         let ys = ys[0].to_owned();

src/models/yolo.rs

@@ -1,6 +1,6 @@
 use anyhow::Result;
 use clap::ValueEnum;
-use image::DynamicImage;
+use image::{DynamicImage, ImageBuffer};
 use ndarray::{s, Array, Axis, IxDyn};
 use regex::Regex;
 
@@ -40,16 +40,16 @@ pub struct YOLO {
 }
 
 impl YOLO {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.batch().to_owned(),
             engine.height().to_owned(),
             engine.width().to_owned(),
         );
 
-        let task = match &options.yolo_task {
-            Some(task) => task.to_owned(),
+        let task = match options.yolo_task {
+            Some(task) => task,
             None => match engine
                 .try_fetch("task")
                 .unwrap_or("detect".to_string())
@@ -60,12 +60,12 @@ impl YOLO {
                 "pose" => YOLOTask::Pose,
                 "segment" => YOLOTask::Segment,
                 "obb" => YOLOTask::Obb,
-                x => todo!("{:?} is not supported for now!", x),
+                x => todo!("Not supported: {x:?} "),
             },
         };
 
         // try from custom class names, and then model metadata
-        let mut names = options.names.to_owned().or(Self::fetch_names(&engine));
+        let mut names = options.names.or(Self::fetch_names(&engine));
         let nc = match options.nc {
             Some(nc) => {
                 match &names {
@@ -88,7 +88,7 @@ impl YOLO {
             },
         };
 
-        let names_kpt = options.names2.to_owned().or(None);
+        let names_kpt = options.names2.or(None);
 
         // try from model metadata
         let nk = engine
@@ -131,10 +131,18 @@ impl YOLO {
 
     pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
         let xs_ = match self.task {
-            YOLOTask::Classify => ops::resize(xs, self.height() as u32, self.width() as u32)?,
-            _ => ops::letterbox(xs, self.height() as u32, self.width() as u32, 114.0)?,
+            YOLOTask::Classify => {
+                ops::resize(xs, self.height() as u32, self.width() as u32, "bilinear")?
+            }
+            _ => ops::letterbox(
+                xs,
+                self.height() as u32,
+                self.width() as u32,
+                "catmullRom",
+                Some(114),
+            )?,
         };
-        let xs_ = ops::normalize(xs_, 0.0, 255.0);
+        let xs_ = ops::normalize(xs_, 0., 255.);
         let ys = self.engine.run(&[xs_])?;
         self.postprocess(ys, xs)
     }
@@ -333,29 +341,35 @@ impl YOLO {
                                 let mask = coefs.dot(&proto).into_shape((nh, nw, 1))?; // (nh, nw, n)
 
                                 // build image from ndarray
-                                let mask_im = ops::build_dyn_image_from_raw(
-                                    mask.into_raw_vec(),
-                                    nw as u32,
-                                    nh as u32,
-                                );
+                                let mask: ImageBuffer<image::Luma<_>, Vec<f32>> =
+                                    match ImageBuffer::from_raw(
+                                        nw as u32,
+                                        nh as u32,
+                                        mask.clone().into_raw_vec(),
+                                    ) {
+                                        Some(buf) => buf,
+                                        None => continue,
+                                    };
+                                let mask = image::DynamicImage::from(mask);
 
-                                // rescale masks
+                                // rescale
                                 let mask_original = ops::descale_mask(
-                                    mask_im,
+                                    mask,
                                     nw as f32,
                                     nh as f32,
                                     image_width,
                                     image_height,
                                 );
-
-                                // crop mask with bbox
                                 let mut mask_original = mask_original.into_luma8();
+
+                                // crop mask
                                 for y in 0..image_height as usize {
                                     for x in 0..image_width as usize {
                                         if x < bbox.xmin() as usize
                                             || x > bbox.xmax() as usize
                                             || y < bbox.ymin() as usize
                                             || y > bbox.ymax() as usize
+                                        // || mask_original.get_pixel(x as u32, y as u32).0 < [127]
                                         {
                                             mask_original.put_pixel(
                                                 x as u32,
@@ -367,23 +381,25 @@ impl YOLO {
                                 }
 
                                 // get masks from image
-                                let mut masks: Vec<Polygon> = Vec::new();
                                 let contours: Vec<imageproc::contours::Contour<i32>> =
                                     imageproc::contours::find_contours_with_threshold(
                                         &mask_original,
-                                        1,
+                                        0,
                                     );
-                                contours.iter().for_each(|contour| {
-                                    if contour.points.len() > 2 {
-                                        masks.push(
-                                            Polygon::default()
-                                                .with_id(bbox.id())
-                                                .with_points_imageproc(&contour.points)
-                                                .with_name(bbox.name().cloned()),
-                                        );
-                                    }
-                                });
-                                y_polygons.extend(masks);
+                                let polygon = match contours
+                                    .iter()
+                                    .map(|x| {
+                                        Polygon::default()
+                                            .with_id(bbox.id())
+                                            .with_points_imageproc(&x.points)
+                                            .with_name(bbox.name().cloned())
+                                    })
+                                    .max_by(|x, y| x.area().total_cmp(&y.area()))
+                                {
+                                    None => continue,
+                                    Some(x) => x,
+                                };
+                                y_polygons.push(polygon);
                             }
                             y = y.with_polygons(&y_polygons);
                         }

src/models/yolop.rs

@@ -15,8 +15,8 @@ pub struct YOLOPv2 {
 }
 
 impl YOLOPv2 {
-    pub fn new(options: &Options) -> Result<Self> {
-        let engine = OrtEngine::new(options)?;
+    pub fn new(options: Options) -> Result<Self> {
+        let mut engine = OrtEngine::new(&options)?;
         let (batch, height, width) = (
             engine.batch().to_owned(),
             engine.height().to_owned(),
@@ -37,8 +37,14 @@ impl YOLOPv2 {
     }
 
     pub fn run(&mut self, xs: &[DynamicImage]) -> Result<Vec<Y>> {
-        let xs_ = ops::letterbox(xs, self.height() as u32, self.width() as u32, 114.0)?;
-        let xs_ = ops::normalize(xs_, 0.0, 255.0);
+        let xs_ = ops::letterbox(
+            xs,
+            self.height() as u32,
+            self.width() as u32,
+            "bilinear",
+            Some(114),
+        )?;
+        let xs_ = ops::normalize(xs_, 0., 255.);
         let ys = self.engine.run(&[xs_])?;
         self.postprocess(ys, xs)
     }
@@ -116,19 +122,19 @@ impl YOLOPv2 {
             let mask_da = mask_da.into_luma8();
             let mut y_polygons: Vec<Polygon> = Vec::new();
             let contours: Vec<imageproc::contours::Contour<i32>> =
-                imageproc::contours::find_contours_with_threshold(&mask_da, 1);
-            contours.iter().for_each(|contour| {
-                if contour.border_type == imageproc::contours::BorderType::Outer
-                    && contour.points.len() > 2
-                {
-                    y_polygons.push(
-                        Polygon::default()
-                            .with_id(0)
-                            .with_points_imageproc(&contour.points)
-                            .with_name(Some("Drivable area".to_string())),
-                    );
-                }
-            });
+                imageproc::contours::find_contours_with_threshold(&mask_da, 0);
+            if let Some(polygon) = contours
+                .iter()
+                .map(|x| {
+                    Polygon::default()
+                        .with_id(0)
+                        .with_points_imageproc(&x.points)
+                        .with_name(Some("Drivable area".to_string()))
+                })
+                .max_by(|x, y| x.area().total_cmp(&y.area()))
+            {
+                y_polygons.push(polygon);
+            };
 
             // Lane line
             let x_ll = x_ll
@@ -150,21 +156,19 @@ impl YOLOPv2 {
             );
             let mask_ll = mask_ll.into_luma8();
             let contours: Vec<imageproc::contours::Contour<i32>> =
-                imageproc::contours::find_contours_with_threshold(&mask_ll, 1);
-            let mut masks: Vec<Polygon> = Vec::new();
-            contours.iter().for_each(|contour| {
-                if contour.border_type == imageproc::contours::BorderType::Outer
-                    && contour.points.len() > 2
-                {
-                    masks.push(
-                        Polygon::default()
-                            .with_id(1)
-                            .with_points_imageproc(&contour.points)
-                            .with_name(Some("Lane line".to_string())),
-                    );
-                }
-            });
-            y_polygons.extend(masks);
+                imageproc::contours::find_contours_with_threshold(&mask_ll, 0);
+            if let Some(polygon) = contours
+                .iter()
+                .map(|x| {
+                    Polygon::default()
+                        .with_id(1)
+                        .with_points_imageproc(&x.points)
+                        .with_name(Some("Lane line".to_string()))
+                })
+                .max_by(|x, y| x.area().total_cmp(&y.area()))
+            {
+                y_polygons.push(polygon);
+            };
 
             // save
             ys.push(

src/ys/bbox.rs

@@ -31,7 +31,7 @@ impl std::fmt::Debug for Bbox {
         f.debug_struct("Bbox")
             .field("xyxy", &[self.x, self.y, self.xmax(), self.ymax()])
             .field("id", &self.id)
-            .field("id_born", &self.id_born)
+            // .field("id_born", &self.id_born)
             .field("name", &self.name)
             .field("confidence", &self.confidence)
             .finish()