HHTseng / video-classification

Tutorial for video classification/ action recognition using 3D CNN/ CNN+RNN on UCF101
940 stars 216 forks source link

cost of CRNN #29

Open Aiiiiii opened 4 years ago

Aiiiiii commented 4 years ago

Hi, I've changed to my own dataset and using the same architecture of your CRNN. For training ONE EPOCH it tales around 30mins, which is incredible long. I used the same dataloader, the image size it 200*200, CNN part is customised as a very basic two layered one. Do you have any suggestions? Thanks!

MounirB commented 4 years ago

Can you tell us about the size of your dataset ? What kind of GPU are you using ?

Aiiiiii commented 4 years ago

Can you tell us about the size of your dataset ? What kind of GPU are you using ?

My dataset has 217,407 instances, 6 frames per video, each frame is 240*240. I am using a TITAN V.