We use the environment same to DINO-DETR, other versions might be available as well. We test our models underpython=3.7,pytorch=1.9.1,cuda=11.1
.
- Install Pytorch and torchvision. Follow the instruction on https://pytorch.org/get-started/locally/.
conda install -c pytorch pytorch torchvision
- Compiling CUDA operators
cd models/EDT/ops
python setup.py build install
-
We have put our model checkpoints hereMMDT-B: Goole Drive.
-
Please download weights and organize them as following:
weights/
├── detect/
├── events/
├── images/
├── checkpoing0011.pth
- Run detect.py and the prediction results in
weights/detect/
.