Joshua Z. Zhang
Joshua Z. Zhang
The old default network is darknet 19, I switched to resnet50 as base model.
src/operator/contrib/yolo_output*
If you run the latest master, stack_neighbor operator shouldn't be required anymore.
stack_neighbor should not exists in latest models anymore, you guys can use the new pre-trained resnet model.
You need to build the mxnet module rather than official one. I am thinking of providing a docker with prebuilt mxnet so it's available out of box
You will need a pretrained model using imagenet, the one provided is in readme
Well, I don't know if they wish to submit PR to mxnet.
What do you mean a python version?
You can checkout a yolov3 version in python: https://gluon-cv.mxnet.io/build/examples_detection/train_yolo_v3.html#sphx-glr-build-examples-detection-train-yolo-v3-py
Wow, thanks for the finding. I've double checked with darknet, the concatenated layers are conv5_5(+1 conv) and stage 7, as in yolo-voc.cfg. Seems like I missed an extra conv layer....