ｈｊｇ

@zhaoyue-zephyrus Hi! Would you please provide the code to generate "normalized_proposal_list.txt" for I find so many people like me are struggling on this problem.

same question Have you worked out yet?

@iuhiyuh Hi！I wonder if you have worked out this problem by SSN? When I run "python gen_proposal_list.py DATASET FRAMES_PATH" an error occurs : "KeyError: 'video_validation_0000893'".

Have solved this problem yet?I have encountered the same problem.

You may put the extracted frames in /mnt/SSD/thumos14/.I fixed the error by trying this.

Hi! Have figure it out yet? I have the same question.

Maybe he means it is extracted after training with ActivityNet-1.3 pretrained model? It will make sense then.

但是生成proposal的基础不就是识别吗

我测试了一下用denseflow处理20个30min的视频，并行多线程处理花费2个多小时，但是单独一个个按顺序处理只需要1个小时出头。

> 我有个问题，BMN训练的时候不是也有类别信息输入，为什么完成提名生成任务后，还是不能确定类别信息呢