Tensorrt-yolo

Breaking Changes

- Add Dockerfile for building project Docker image (73060db40af205aeb2fd81d144294f2efd09f95f)
- Add support for YOLOv3 and Ultralytics model export (8f2af943b058cd143647658acdcc26eb31648503)
- Added Support for YOLOv10 (62071cb0993b87220243b8f9aaa35fe98b6917d1)
- Refactor deploy library (8a4df3369e6e35fec2ce0baf8a3758457f45642f)
- Use CUDA Graph to Accelerate Static Model Inference (3576c78d5063a8d1a00544a9e0472b4a8af45389)
- Refactor: Add BaseDet base class, refactor DeployDet and DeployCGDet (416e77b64748d3f69159e6674ea13b90283c3716)
- Add streaming video support using VideoPipe (https://github.com/laugh12321/TensorRT-YOLO/issues/17, https://github.com/laugh12321/TensorRT-YOLO/issues/19)
- feat: major update with pybind11 integration and new 4.0 tag (3244eeaf69b69a5db7d7f6458cfe8e707809c70e)

Bug Fixes

- Fix incorrect time interval calculation (a3dee2a16f2ac2635970f5c68d168de292768b03)
- Fix: Include `<cstring>` to resolve "memcpy is not a member of std" error in Linux (1b763d97bed8eb3e19b70e4aa96a497149a3efed)
- Update detection output variable names for clarity (fe67b018e946b18a123f1b0b4468cb28abcd5d39)
- Add `cstdint` to resolve Linux compilation issue (f602dc424547e66e7bffe747def81c31f6a61519)
- Fix: Graph input and output tensors must include dtype information. (2cacb7d1cc3dbbc4cb8a0ddc581a2cccc7a6b783)

**Full Changelog**: https://github.com/laugh12321/TensorRT-YOLO/compare/v3.0...v4.0

Breaking Changes

- Add TensorRT INT8 PTQ support (87f67ffbfffe002cbed5d1a42afd0f55f744faa2)
- Add C++ inference implementation (0f3069f8955e3ce2f20be5a85498efded1e8aeec)
- Implemented parallel preprocessing with multiple streams (86d6175a21c9ed87d1dcf46ef65e18e3f8b5cd5a)
- Refactor C++ inference code to support dynamic and static libraries (425a1a48e2c3d56529837951eddd98c0b89c170a)
- Refactored Python code related to TensorRT-YOLO and packaged it as tensorrt_yolo. (a10ebc87973b3fa925d4bc1d7e8d4dc397d500c1)

Bug Fixes

- Fix batch visualize bug (9125219ea8ea02adf16d3a3fd4d16bbb056a50b4)
- Remove deleted move constructor and move assignment operator (e287342884285d406a0985d6ec03f13475b2a9b9)
- Fix duplicate imports (1237e21c458c43e00aa7fd3938fd9d42b91b4cde)
- Fix bug (24ea950b31f6c73ff9bd853034310da45d45c02b)

**Full Changelog**: https://github.com/laugh12321/TensorRT-YOLO/compare/v2.0...v3.0

Breaking Changes

- Implement YOLOv9 Export to ONNX and TensorRT with EfficientNMS Plugin (249bfab3e5ca59d6ab3a05491955fcf26adaf6ae)
- Remove FLOAT16 ONNX export and add support for Dynamic Shape export (9ec1f291f9fb417827b10228c6c7cb8f2e944f43)
- Enable dynamic shape inference with CUDA Python and TensorRT 8.6.1 for inference (328645029f3b86c1f733e35c269b6cd673f88212)

Bug Fixes

- Fix bounding boxes rescales bug (0f90cd0a77117b7923a7c4c8b70b828bccf153f6)
- Fix AttributeError in YOLOv9 Model Export (8)

**Full Changelog**: https://github.com/laugh12321/TensorRT-YOLO/compare/v1.0...v2.0

Breaking Changes

- Supports FLOAT32, FLOAT16 ONNX export, and TensorRT inference
- Supports YOLOv5, YOLOv8, PP-YOLOE, and PP-YOLOE+
- Integrates EfficientNMS TensorRT plugin for accelerated post-processing
- Utilizes CUDA kernel functions to accelerate preprocessing
- Supports Python inference

Bug Fixes

- Fix pycuda.driver.CompileError on Jetson (1)
- Fix Engine Deserialization Failed using YOLOv8 Exported Engine (2)
- Fix Precision Anomalies in YOLOv8 FP16 Engine (3)
- Fix YOLOv8 EfficientNMS output shape abnormality (0e542ee732b176590732fa013693cfc2417a8c5c)
- Fix trtexec Conversion Failure for YOLOv5 and YOLOv8 ONNX Models on Linux) (4)
- Fix Inference Anomaly Caused by preprocess.cu on Linux (5)

**Full Changelog**: https://github.com/laugh12321/TensorRT-YOLO/commits/v1.0