You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, I implement a SSD detector based on torchvision's faster-rcnn, named demonet. I use similar structure with torchvison's. Currently I use mobilenet_v2 as the backbone.
Now I focus on model's inference, to make it have fully support for torchscript and onnx. I borrow a lot of torchvision's implementation as my code's cornerstone, so it is very easy to port if you are familiar with torchvision's. DEMONET have support for torchscript now, and I will try the onnx export. You can check some of my code snippets, generalized_ssd.py and box_head.py.
I only train demonet on pascal voc datasets, and with map about 66.1, so there is a lot to do to get a Good pretrained model. Maybe you can try it.
Please add other detectors like SSD and SSDlite.
The text was updated successfully, but these errors were encountered: