why don't you use flownet2 for generating optical flow?

StevenLiuWen / ano_pred_cvpr2018

Official implementation of Paper Future Frame Prediction for Anomaly Detection -- A New Baseline, CVPR 2018

435 stars 126 forks source link

why don't you use flownet2 for generating optical flow? #9

Closed owliz closed 6 years ago

owliz commented 6 years ago

Hi, Steven. Thanks for your inspiring paper and code.
I want to know why you choose flownet-SD rather than flownet2(the final fused model) to generate optical flow. Did flownet2 show poor performance in your experiments?
Do you think a more accurate optical flow prediction will improve the final detection results?

Looking forward to your insight.

StevenLiuWen commented 6 years ago

FlowNet2 is an ensemble model (combine, not fully-differentiable, it can not do