0.0.5 (#24)

* Using Rayon to accelarate YOLO post-processing * Refactor YOLO with outputs format * Optimize `conf * clss` for yolov5 v6 v7 * Add depth-anything-v2 * Update README.md * Update CHANGELOG.md
jamjamjon · Jul 12, 2024 · edc3a88 · edc3a88
1 parent 25d9088
commit edc3a88
Show file tree

Hide file tree

Showing 69 changed files with 1,564 additions and 1,204 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,20 +1,46 @@
+## v0.0.5 - 2024-07-12
+
+### Changed
+
+- Accelerated `YOLO`'s post-processing using `Rayon`. Now, `YOLOv8-seg` takes only around **~8ms (~20ms in the previous version)**, depending on your machine. Note that this repo's implementation of `YOLOv8-Segment` saves not only the masks but also their contour points. The official `YOLOv8` Python version only saves the masks, making it appear much faster.
+- Merged all `YOLOv8-related` solution models into YOLO examples.
+- Consolidated all `YOLO-series` model examples into the YOLO example.
+- Refactored the `YOLO` struct to unify all `YOLO versions` and `YOLO tasks`. It now supports user-defined YOLO models with different `Preds Tensor Formats`.
+- Introduced a new `Nms` trait, combining `apply_bboxes_nms()` and `apply_mbrs_nms()` into `apply_nms()`.
+
+### Added
+
+- Added support for `YOLOv6` and `YOLOv7`.
+- Updated documentation for `y.rs`.
+- Updated documentation for `bbox.rs`.
+- Updated the `README.md`.
+- Added `with_yolo_preds()` to `Options`.
+- Added support for `Depth-Anything-v2`.
+- Added `RTDETR` to the `YOLOVersion` struct.
+
+### Removed
+
+- Merged the following models' examples into the YOLOv8 example: `yolov8-face`, `yolov8-falldown`, `yolov8-head`, `yolov8-trash`, `fastsam`, and `face-parsing`.
+- Removed `anchors_first`, `conf_independent`, and their related methods from `Options`.
+
+
 ## v0.0.4 - 2024-06-30
 
 ### Added
 
-- Add X struct to handle input and preprocessing  
+- Add X struct to handle input and preprocessing
 - Add Ops struct to manage common operations
 - Use SIMD (fast_image_resize) to accelerate model pre-processing and post-processing.YOLOv8-seg post-processing (~120ms => ~20ms), Depth-Anything post-processing (~23ms => ~2ms).
 
 ### Deprecated
-- Mark `Ops::descale_mask()` as deprecated.
 
+- Mark `Ops::descale_mask()` as deprecated.
 
-### Fixed 
+### Fixed
 
 ### Changed
 
-### Removed 
+### Removed
 
 ### Refactored
 

diff --git a/Cargo.toml b/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "usls"
-version = "0.0.4"
+version = "0.0.5"
 edition = "2021"
 description = "A Rust library integrated with ONNXRuntime, providing a collection of ML models."
 repository = "https://github.com/jamjamjon/usls"
@@ -11,7 +11,7 @@ exclude = ["assets/*", "examples/*"]
 
 [dependencies]
 clap = { version = "4.2.4", features = ["derive"] }
-ndarray = { version = "0.15.6" }
+ndarray = { version = "0.15.6", features = ["rayon"] }
 ort = { version = "2.0.0-rc.2", default-features = false, features = [
     "load-dynamic",
     "copy-dylibs",

diff --git a/README.md b/README.md
diff --git a/assets/falldown.jpg b/assets/falldown.jpg
diff --git a/assets/trash.jpg b/assets/trash.jpg
diff --git a/examples/depth-anything/README.md b/examples/depth-anything/README.md
@@ -9,6 +9,7 @@ cargo run -r --example depth-anything
 - [depth-anything-s-dyn](https://github.com/jamjamjon/assets/releases/download/v0.0.1/depth-anything-s-dyn.onnx)  
 - [depth-anything-b-dyn](https://github.com/jamjamjon/assets/releases/download/v0.0.1/depth-anything-b-dyn.onnx)
 - [depth-anything-l-dyn](https://github.com/jamjamjon/assets/releases/download/v0.0.1/depth-anything-l-dyn.onnx)
+- [depth-anything-v2-s](https://github.com/jamjamjon/assets/releases/download/v0.0.1/depth-anything-v2-s.onnx)  
 
 
 ## Results

diff --git a/examples/depth-anything/main.rs b/examples/depth-anything/main.rs
@@ -1,9 +1,10 @@
 use usls::{models::DepthAnything, Annotator, DataLoader, Options};
 
 fn main() -> Result<(), Box<dyn std::error::Error>> {
-    // visual
+    // options
     let options = Options::default()
-        .with_model("depth-anything-s-dyn.onnx")?
+        // .with_model("depth-anything-s-dyn.onnx")?
+        .with_model("depth-anything-v2-s.onnx")?
         .with_i00((1, 1, 8).into())
         .with_i02((384, 512, 1024).into())
         .with_i03((384, 512, 1024).into());

diff --git a/examples/face-parsing/README.md b/examples/face-parsing/README.md
diff --git a/examples/face-parsing/main.rs b/examples/face-parsing/main.rs
diff --git a/examples/fastsam/README.md b/examples/fastsam/README.md
diff --git a/examples/fastsam/main.rs b/examples/fastsam/main.rs
diff --git a/examples/rtdetr/README.md b/examples/rtdetr/README.md
diff --git a/examples/rtdetr/demo.png b/examples/rtdetr/demo.png
diff --git a/examples/rtdetr/main.rs b/examples/rtdetr/main.rs
diff --git a/examples/yolo-world/README.md b/examples/yolo-world/README.md
diff --git a/examples/yolo-world/main.rs b/examples/yolo-world/main.rs