代码中的一些可能bug?

[long2double]

两个问题

第一个问题： multitracker.py中update_sep3函数中已用F.normalize, with torch.no_grad(): output = self.model(im_blob)[-1] hm = output['hm'].sigmoid_() wh = output['wh'] id_feature = output['id'] id_feature = F.normalize(id_feature, dim=1) 然而，STrack类中，update_features方法中，再次使用np.linalg.norm(feat)是否已经重复归一化操作？（需要去掉F.normalize???）

第二个问题： update_sep3函数中，504行 for i_class in range(1, self.opt.num_classes + 1): activated_stracks = [] refind_stracks = [] lost_stracks = [] removed_stracks = [] dets = dets_classes[i_class] if conf_thres is None: remain_inds = dets[:, 4] > self.opt.conf_thres else: remain_inds = dets[:, 4] > conf_thres[i_class-1] dets = dets[remain_inds] ############是否少了索引remain_inds?????############### id_feature = cls_id_feats[i_class-1]

[dingwoai]

Indeed!!! Thx a lot!

For the second one, did you get better results after fixing this?

[long2double]

暂时还没有进行训练，只是看了下源码，后续有结果在联系