The training and validation files for the PKU dataset

[li-stu-998]

Hello, could you please provide the training and validation files for the PKU dataset, or the code for generating the training and validation files?

[cong-wu]

Hi,
The code for generating training and validation files is basically similar to data_gen/ntu_gendata.py. You can make it suitable for pku dataset with the reference of 'https://github.com/Levigty/AimCLR/tree/main/tools'.

[li-stu-998]

Hi, The code for generating training and validation files is basically similar to data_gen/ntu_gendata.py. You can make it suitable for pku dataset with the reference of 'https://github.com/Levigty/AimCLR/tree/main/tools'.

Hello, thank you for providing the link. According to the above link you provided, I generated "train_data. npy" and "train_num_frame. npy" for PKUMMD-I. However, when I trained according to the configuration in your paper, I was unable to train successfully and the recognition accuracy did not improve. May I ask if the sizes of these two files are "3.16GB" and "147KB" respectively? Is the learning rate of your experimental configuration 0.01 and the encoder k 8192?

[Mahmoud-Ali-FCIS]

Hello, Thanks for your work @cong-wu , I have the same problem of PKUMMD-I & II. I cant reproduce your result in the paper after following your configurarion. I used the mentioned link for data generation "train_data. npy" and "train_num_frame. npy".
Could you please share the preprocess data for PKU?
@li-stu-998 Did you manage to solve the problem?
Thanks.