search

IDEA-Research / DINO

[ICLR 2023] Official implementation of the paper "DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection"
Apache License 2.0
2.24k stars 250 forks source link

Plan to support for ONNX #100

Open jhj7905 opened 2 years ago

jhj7905 commented 2 years ago

These days, I trained and evaluated my own dataset and got significant result with your model compared to YOLOS. Now, I have tried to convert DINO(pytorch) into DINO(onnx or tensorrt). But there are a few reference. so still in trouble. Do you have a plan to support for onnx? To all DINO contributors, thank you for sharing the code!

lucasjinreal commented 2 years ago

there were some exploration in converting DETR to onnx, hope it help. https://github.com/jinfagang/yolov7_d2 Haven't try DINO though, but am trying porting DINO to detectron2 these days, converting transformer based model to onnx a little bit cubersome, even you can export, probabely not work on TensorRT, but we have made it work on DETR already.

jhj7905 commented 2 years ago

@jinfagang Thank you for sharing the reference. i will try to apply it to convert DETR(pytorch) to onnx(or tenssorrt)

yangzheng0218 commented 1 year ago

@jinfagang Thank you for sharing the reference. i will try to apply it to convert DETR(pytorch) to onnx(or tenssorrt)

Is your conversion successful?

lonalala commented 11 months ago

Did you successfully converted DINO to ONNX? thank you