chenguitao

https://github.com/shenweichen/DeepCTR/blob/master/deepctr/inputs.py#L41 若是多列特征共用同个字典的话，share_embed，设置feat.embedding_name参数的话，需要将https://github.com/shenweichen/DeepCTR/blob/master/deepctr/inputs.py#L41修改为name=prefix + '_seq_emb_' + feat.embedding_name,

TSN model was trained on Kinetics datase with Resnet50

1

hello , If you have the model, would you like to share it, thank you very much

load model error

1

KeyError: 'unexpected key "pretrained.conv1.1.num_batches_tracked" in state_dict' please tell me which error for this? thank you

watermark removal mask in video

3

This is really an exciting project. Thank you for sharing. I want to delogo use it for video, watermark removal in your demo video. I want to ask how the...

四层特征图进行融合得到特征图用F表示

1

你代码中做了5层特征融合，将P2，P3，P4, P5通通上采样到P1大小然后concat，与论文有所不同，是你故意如此修改的吗？

hello! you get comparable results include RGB model, flow model, fusion model, but eval_net.py only generate RGB or flow result, where fusion result and fusion model? thank you ,brother

1

Sorry, I do not understand what you said. _Originally posted by @shuangshuangguo in https://github.com/shuangshuangguo/caffe2pytorch-tsn/issues/10#issuecomment-442074846_

gen_concated_feature

1

你好，我觉得gen_concated_feature.py文件中，sparse_slots_filter = [slot_name for slot_name in slots_filter if slot_name in self.sparse_feats_slots] 应该改为sparse_slots_filter = [slot_name for slot_name in slots_filter if slot_name in (self.sparse_feats_slots+self.list_sparse_feats_slots)]

bother, if you have it , could you share? thank you

Brother, you later found the pre-training model about TSN on Kinetics datase with Resnet50 ? If you have it, would you like to share it, thank you very much!