About the difference between fg_depth and depth

[sinsin1998]

fg_info = img_metas[sample_idx]['foreground2D_info']
fg_pxl = fg_info['fg_pixels'][view_idx]
fg_depth = torch.from_numpy(fg_pxl[:,2]).to(device) in line 207 in MSMDFusionDetector
and
fg_real_pixels = img_metas[i]['foreground2D_info']['fg_real_pixels']
depth = fg_real_pxl[:,2]

code above shows that you have use two depth information in the generation of foreground2D, could you tell me the deffrence between them ?

[SxJyJay]

"fg_pxl" corresponds to the location of seeds sampled in MDU. "fg_real_pixels" corresponds to the locations of projected lidar points. Here, locations are represented in the form of (u, v, d), where u and v are pixel coordinates and d is the depth.

[Libraaer]

fg_info = img_metas[sample_idx]['foreground2D_info'] fg_pxl = fg_info['fg_pixels'][view_idx] fg_depth = torch.from_numpy(fg_pxl[:,2]).to(device) in line 207 in MSMDFusionDetector and fg_real_pixels = img_metas[i]['foreground2D_info']['fg_real_pixels'] depth = fg_real_pxl[:,2]

code above shows that you have use two depth information in the generation of foreground2D, could you tell me the deffrence between them ?

Hello, have you trained this model? I'm using A800 here but still not enough to train. If you've trained, can you give me an answer to the reason for my lack of memory here, thank you