Xharlie / pointnerf

Point-NeRF: Point-based Neural Radiance Fields
Other
1.1k stars 127 forks source link

CodeErroe #101

Open iszhihao opened 8 months ago

iszhihao commented 8 months ago

In this code, are points_color_all and points_dir_all reversed when using torch.cat?

if points_xyz_all is not None: if opt.bgmodel.startswith("planepoints"): gen_pnts, gen_embedding, gen_dir, gen_color, gen_conf = train_dataset.get_plane_param_points() visualizer.save_neural_points("pl", gen_pnts, gen_color, None, save_ref=False) print("vis pl") points_xyz_all = torch.cat([points_xyz_all, gen_pnts], dim=0) points_embedding_all = torch.cat([points_embedding_all, gen_embedding], dim=1) points_color_all = torch.cat([points_color_all, gen_dir], dim=1) points_dir_all = torch.cat([points_dir_all, gen_color], dim=1) points_conf_all = torch.cat([points_conf_all, gen_conf], dim=1) model.set_points(points_xyz_all.cuda(), points_embedding_all.cuda(), points_color=points_color_all.cuda(), points_dir=points_dir_all.cuda(), points_conf=points_conf_all.cuda(), Rw2c=normRw2c.cuda() if opt.load_points < 1 and opt.normview != 3 else None) epoch_count = 1 total_steps = 0 del points_xyz_all, points_embedding_all, points_color_all, points_dir_all, points_conf_all