Architecture | Documentation | Blog | Roadmap | Slack
The repository mainly provides 2D model inference functionality, and the code provides daily development of packaged libs for integration, testing, and inference. The framework provides multi-threaded, singleton pattern, producer and consumer patterns. Cache log analysis is also supported.
| Libraries | Eigen | Gflags | Glog | Yaml-cpp | Cuda | Cudnn | Tensorrt | Opencv |
|---|---|---|---|---|---|---|---|---|
| Version | 3.4 | 2.2.2 | 0.6.0 | 0.8.0 | 11.4 | 8.4 | 8.4 | 3.4.5 |
Visit our documentation to learn more.
- Dataset: The validation dataset is TinyCOCO, which contains 1,000 training samples and 500 test samples. All models in the table were trained on the full COCO2017 dataset.
- Model: The deployed model is the 's' version of the YOLO series.
- Quantize: Quantization was performed using NVIDIA's Post-Training Quantization (PTQ) method.
| Model | Platform | Resolution | mAP50-95(fp32) | mAP50(fp32) | mAP50-95(fp16) | mAP50(fp16) | mAP50-95(int8) | mAP50(int8) |
|---|---|---|---|---|---|---|---|---|
| Yolov5s | RTX4060/orin x | 640x640 | 0.458 | 0.619 | 0.459 | 0.619 | 0.424 | 0.576 |
| Yolov8s | RTX4060/orin x | 480x640 | 0.467 | 0.622 | 0.468 | 0.622 | 0.453 | 0.604 |
| Yolov11s | RTX4060/orin x | 480x640 | 0.491 | 0.656 | 0.491 | 0.657 | 0.458 | 0.607 |
| Yolox | RTX4060/orin x | 416x416 | - | - | - | - | - | - |
Welcome users to participate in these projects. Please refer to CONTRIBUTING.md for the contributing guideline.We encourage you to join the effort and contribute feedback, ideas, and code. You can participate in Working Groups, Working Groups have most of their discussions on Slack or QQ (938558640).