Question on seq_len

[Tao-11-chen]

Hi, i'm bit confused on the seq_len here, it seems like a fixed value(16), so if i input a video less than 16 frames, it will not be processed?

line 606 in point_tracker_3d.py

     for window_end in range(self.seq_len, T + 1, self.seq_len // 2):
         window_start = window_end - self.seq_len

        coords_init = pred.coords[:, window_start : window_start + self.seq_len // 2] # (B, W/2, N, 3)
        visibs_init = pred.visibs[:, window_start : window_start + self.seq_len // 2] # (B, W/2, N)
        coords_init = torch.cat([coords_init, repeat(coords_init[:, -1], "b n c -> b w n c", w=self.seq_len // 2)], dim=1)
        visibs_init = torch.cat([visibs_init, repeat(visibs_init[:, -1], "b n -> b w n", w=self.seq_len // 2)], dim=1)