od_opencv 0.8.0

Object detection utilities in Rust programming language for YOLO-based neural networks in OpenCV ecosystem
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110

//! Object detection utilities for YOLO-based neural networks.
//!
//! This crate provides wrappers for running YOLO models.
//!
//! ## Backends
//!
//! - `opencv-backend` (default): Uses OpenCV DNN for inference. Supports Darknet and ONNX models.
//! - `ort-backend`: Uses ONNX Runtime for inference. Pure Rust, no OpenCV required.
//!
//! ## Supported Models
//!
//! - Traditional YOLO (v3, v4, v7) - Darknet format, opencv-backend only
//! - YOLOv5 - ONNX format, both backends
//! - Ultralytics YOLO (v8, v9, v11) - ONNX format, both backends

// Common types (always available)
pub mod bbox;
pub mod image_buffer;
pub mod model_trait;
pub mod model_factory;

pub use bbox::BBox;
pub use image_buffer::{ChannelOrder, ImageBuffer};
pub use model_trait::ObjectDetector;
pub use model_factory::Model;

// Pure Rust preprocessing/postprocessing (for ort-backend, rknn-backend, tensorrt-backend)
#[cfg(any(feature = "ort-backend", feature = "rknn-backend", feature = "tensorrt-backend"))]
pub mod preprocessing;

#[cfg(any(feature = "ort-backend", feature = "rknn-backend", feature = "tensorrt-backend"))]
pub mod postprocess;

// OpenCV DNN backend - requires opencv/dnn feature
#[cfg(feature = "opencv-backend")]
pub mod backend_opencv;

// ONNX Runtime backend
#[cfg(feature = "ort-backend")]
pub mod backend_ort;

// RKNN NPU backend
#[cfg(feature = "rknn-backend")]
pub mod backend_rknn;

// TensorRT backend
#[cfg(feature = "tensorrt-backend")]
pub mod backend_tensorrt;

// DNN backend/target enums for opencv-backend
#[cfg(feature = "opencv-backend")]
pub mod dnn_backend;

#[cfg(feature = "opencv-backend")]
pub use dnn_backend::{DnnBackend, DnnTarget};

// Backwards-compatible re-exports for opencv-backend
// which allows existing code using
// `od_opencv::model_ultralytics::...` to still work
#[cfg(feature = "opencv-backend")]
pub use backend_opencv::utils;

#[cfg(feature = "opencv-backend")]
pub use backend_opencv::model_format;

#[cfg(feature = "opencv-backend")]
pub use backend_opencv::model_classic;

#[cfg(feature = "opencv-backend")]
pub use backend_opencv::model_ultralytics;

#[cfg(feature = "opencv-backend")]
pub use backend_opencv::model_yolov5;

// model module is needed for ModelTrait (opencv-backend only - depends on dnn)
#[cfg(feature = "opencv-backend")]
pub use backend_opencv::model;

// Re-exports for ort-backend
#[cfg(feature = "ort-backend")]
pub use backend_ort::ModelUltralyticsOrt;

#[cfg(feature = "ort-backend")]
pub use backend_ort::ModelYOLOv5Ort;

#[cfg(feature = "ort-backend")]
pub use backend_ort::OrtModelError;

// Re-exports for rknn-backend
#[cfg(feature = "rknn-backend")]
pub use backend_rknn::ModelUltralyticsRknn;

#[cfg(feature = "rknn-backend")]
pub use backend_rknn::RknnModelError;

// Re-exports for tensorrt-backend
#[cfg(feature = "tensorrt-backend")]
pub use backend_tensorrt::ModelUltralyticsRt;

#[cfg(feature = "tensorrt-backend")]
pub use backend_tensorrt::TrtModelError;

// OpenCV compatibility layer for ORT backend
// Allows using ORT inference with OpenCV Mat input
// Provides ModelTrait that does NOT depend on opencv/dnn
#[cfg(feature = "ort-opencv-compat")]
pub mod opencv_compat;

#[cfg(feature = "ort-opencv-compat")]
pub use opencv_compat::{mat_to_array_view, mat_to_array3, preprocess_mat, ModelTrait};