search

liangchen527 / SLADD

Official code for Self-supervised Learning of Adversarial Example: Towards Good Generalizations for Deepfake Detection (CVPR 2022 oral)
Apache License 2.0
131 stars 16 forks source link

What is the recommended number of frames when training on ResNet18 #8

Closed AmiraAlsamawi closed 1 year ago

AmiraAlsamawi commented 1 year ago

@liangchen527 I am facing a shortage of resources and will train the ResNet18 model instead of Xception. What is your recommendation for the minimum number of frames (samples) per video for the training set? The original work used 270 frames/video on the Xception model, but I want to minimize this number as much as possible.

liangchen527 commented 1 year ago

Sorry, I didn't try other options. Guess the more frames the better?