Open jlimsf opened 3 years ago
You can warp source image, or use optical flow visualization tools https://github.com/tomrunia/OpticalFlow_Visualization
what did you use for your paper to get the visualization? here is what I am trying with flowvis
flow = motion_params['optical_flow'][0].squeeze()
print (flow.shape)
flow_np = flow.detach().cpu().numpy()
flow_im = flow_vis.flow_to_color(flow_np, convert_to_bgr=False)
print (flow_im)
flow_im = Image.fromarray(flow_im)
flow_im.save('flow.png')
this is the output I am getting.
I don't remember. Guess you need to subtract identity meshgrid and renormalize.
How can we view the pixel wise flow in Figure 2? The output is of shape
batch x 64 x 64 x 2