rohitgirdhar / ActionVLAD

ActionVLAD for video action classification (CVPR 2017)
https://rohitgirdhar.github.io/ActionVLAD/
Other
216 stars 61 forks source link

question related to issue 19 #20

Closed theuselesscorporation closed 6 years ago

theuselesscorporation commented 6 years ago

Hi Rohit!! thanks for your response.

For training the ReadMe says to run the script 001_VGG_RGB_HMDB_netvlad_stage1.sh for example. But to run the script I need to have 'imnet_conv5_kmeans64.pkl' file.

I would like to know how can I generate my own file, say for another dataset?

Thanks

rohitgirdhar commented 6 years ago

As I pointed out in that issue, the link is in the get_models.sh: https://cmu.box.com/shared/static/9rq9gy0sjwyr2wxm4kt1kd4ogbzpts77.pkl

Please see the training instructions on how to generate it on your own cluster centers. (001_VGG_RGB_HMDB_netvlad_feats_for_clustering.sh and 001_VGG_RGB_HMDB_netvlad_cluster.sh)