search

daniilidis-group / EV-FlowNet

Code for the paper "EV-FlowNet: Self-Supervised Optical Flow for Event-based Cameras"
Other
169 stars 44 forks source link

Frequency of prediction #6

Open eugenelyj opened 5 years ago

eugenelyj commented 5 years ago

Thanks for your sharing of this work!! I have tested on my PC and also read the paper, it seems that the frequency of optical flow prediction is the same as the RGB frame, right? If so, have you compared it with the work using RGB frame?

alexzzhu commented 5 years ago

Yes, you are correct. We report numbers comparing this method to UnFlow, which is a self-supervised method trained on images, applied to the grayscale frames from the DAVIS.

Jee-King commented 4 years ago

@eugenelyj @alexzzhu Hi, I read this paper. it seems that at training time, the network is provided with the event timestamp images, as well as a pair of grayscale images for calculating optical flow. However, why this test code only uses gray images as input(the test dataset _outdoor1 only contains gray images )? Wainting for your apply. Thanks!

alexzzhu commented 4 years ago

@Jee-King I'm not sure where you are seeing that the test code only uses grayscale images. In the forward pass of the network, here: https://github.com/daniilidis-group/EV-FlowNet/blob/cf801ef5e4b112e08bfe6eac4f0878b1f5aac2fe/src/test.py#L33, the network only sees the event images.