README.md

RF-DETR: SOTA Real-Time Object Detection Model

How to Run

1. set up a virtual environment.

   git clone https://github.com/lyuwenyu/RT-DETR.git
   cd RT-DETR
   conda create -n rtdetr -y python=3.11
   conda activate rtdetr
   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu129
   pip install opencv-python
   pip install onnx
   pip install onnxscript
   pip install onnxsim
   pip install scikit-image
   pip install PyYAML
   pip install tensorboard
   pip install pycocotools
   pip install faster_coco_eval
   

2. download pretrained checkpoints.

   cd rtdetrv2_pytorch
   mkdir -p pretrained
   wget https://github.com/lyuwenyu/storage/releases/download/v0.2/rtdetrv2_r18vd_120e_coco_rerun_48.1.pth -P pretrained
   cd ..
   

3. check pytorch model inference

   cd ..
   python infer.py
   

How to Run (TensorRT)

1. generate onnx file

   python onnx_export_new_ln.py
   

2. generate tensorrt model

   python onnx2trt.py
   

• fp16 (RT-DETRv2-S)
  [TRT_E] 1000 iterations time: 5.0255 [sec]
  [TRT_E] Average FPS: 198.99 [fps]
  [TRT_E] Average inference time: 5.03 [msec]
  GPU mem : 228M

• DETRs Beat YOLOs on Real-time Object Detection

• RT-DETR official GitHub