Yang Fu
Yang Fu
It would be great to add support for pose tracking dataset i.e. posetrack2017/2018.
@innerlee If possible, adding both data loading and tracking part would be great. However, the tracking part seems a little bit complicate and always comes with some extra modules. Maybe,...
Sorry for the late reply, I was really busy. Your understanding is right, we use re-rank between source feature and target feature. The motivation is to obtain a better re-rank...
Maybe you can try "mkdir random_split" first. The "random" means randomly select samples from whole dataset to annotate, while the "cluster" means randomly select one sample from each cluster.
I update the source_train.py, you can try it now.
Sorry I don't know the reasons, can you reproduce the performance using the provide models? You need use provided model as pre-trained model and make sure the num-split is two.
If you cannot load the pretrained model, this [link](https://github.com/CSAILVision/places365/issues/25#issuecomment-333871990) maybe helpful. And please make sure you train our codes on Two GPUs.
@geyutang @Alan-Paul There're some suggestions. First, check if the performance of our provide model is same as reported in the paper. Also, check the performance of pretrained model, which should...
I trained model again with Pytorch=0.4.1, then I got the adaptation result from Market to Duke, as 53.3/72.4(mAP/R1), it's almost the same with the results reported in the paper.