Chengxu Liu

chack your object name, whether they have the Upper letter : you can change "name = obj.find('name').text.lower().strip()" to "name = obj.find('name').text.strip()" in voc_dateset.py

I cannot download VGG_imagenet.npy too.if you have downloaded or got it. Can you sent the VGG_imagenet.npy to my e-mail? ? ? thanks emmmm~~~ my email [email protected]

These are the two different settings in REDS, and the results are of course different. Btw, as far as I know, people usually use _train_shape_ and _train_shape_bicubic_.

the path of result_dir is obtained by the "evaluate_motchallenge.py" and it can output the tracking result and written as ".txt" file.