In the code loss/flow.py
line 35
“param event_list: [batch_size x N x 4] input events (y, x, ts, p) ”
the input seems to be (ts,y, x, p) rather than (y, x, ts, p)
line 44-46
“ flow_idx = event_list[:, :, 1:3].clone()
flow_idx[:, :, 0] = self.res[1] # torch.view is row-major
flow_idx = torch.sum(flow_idx, dim=2)
”
does it mean flow_idx = 240y+x ? I don't know if I misunderstood
In the code loss/flow.py line 35 “param event_list: [batch_size x N x 4] input events (y, x, ts, p) ”
the input seems to be (ts,y, x, p) rather than (y, x, ts, p)
line 44-46 “ flow_idx = event_list[:, :, 1:3].clone() flow_idx[:, :, 0] = self.res[1] # torch.view is row-major flow_idx = torch.sum(flow_idx, dim=2) ” does it mean flow_idx = 240y+x ? I don't know if I misunderstood