I want to use the trained model you released to predict some other videos,
There is a problem that the google feature extraction part cost too much time.
If I need to extract 300 frame features from videos, it needs to 300 times run google inception model.
Batch feature extraction can save much time because I can set batch_size equal to 50 and run the inception model only 6 times.
I tried to use that way but I found that the batch size of the .pb model has been fixed to one.
Do you know some trick or black technology to solve that problem?
Something like re-create a pb model that support batch prediction...
Hello, @miha-skalic , nice work!
I want to use the trained model you released to predict some other videos, There is a problem that the
google feature extraction part
cost too much time. If I need to extract 300 frame features from videos, it needs to 300 times run google inception model. Batch feature extraction can save much time because I can set batch_size equal to 50 and run theinception
model only 6 times.I tried to use that way but I found that the batch size of the .pb model has been fixed to one.
Do you know some trick or black technology to solve that problem? Something like re-create a pb model that support batch prediction...