youjiangxu / seqvlad-pytorch Goto Github PK

Hi,
when I ran your code, I ran into the following problem:

File "/home/lj2017/seqvlad-pytorch/dataset.py", line 180, in getitem
return self.get(record, segment_indices)
File "/home/lj2017/seqvlad-pytorch/dataset.py", line 188, in get
seg_imgs = self._load_image(record.path, p)
File "/home/lj2017/seqvlad-pytorch/dataset.py", line 64, in _load_image
mclient = mc.MemcachedClient.GetInstance(self.server_list_config_file, self.client_config_file)
NameError: name 'mc' is not defined

But, at the beginning of the file (dataset.py) , there are

#sys.path.insert(0, "/mnt/lustre/share/pymc")
#import mc

so, what is 'mc'? thankyou.

Could you provide us the .npz score file for ucf101, just like the corresponding score files for hmdb51 you have provided in this repository shown as follows:

What I need is the .npz score file for ucf01(flow), I am grateful for you kind help, thank you very much.

Hi, glad to meet you, my question is how many segments do you use during training process, 3 segments or 7 segments, thank you very much.

In the code, your model choices include ['seqvlad', 'bidirect', 'unshare_bidirect'], which is the choice you made in the experiment to obtain the best accuracy. Thank you very much.

How did you fuse the iDT based scores with the scores of your method, thank you so much. Looking forward to your answer.

师兄，您好！
我是河北工业大学16级计算机研究生，也在天津。有意向读您导师韩亚洪教授的博士，希望师兄能够给些建议。
本人qq邮箱
[email protected]

Thank you so much for your kind help. My question is do you have the test file before renamed by AllVideos_renamed.txt? This is very important for our experiment, thank you a lot for your kind help. Thank you again for all the helps you have provided, we deeply appreciate it.
Reference:
https://github.com/ZealACMer/seqvlad-pytorch/tree/master/data/hmdb51_splits

I notice that you use a large batchsize=64, so I want to know more details about the GPUs you use.