search

mellody11 / RandLA-Net-Pytorch-New

Pytorch version of RandLA-Net for S3DIS and Semantickitti
MIT License
31 stars 4 forks source link

visualization #13

Open jia1047 opened 5 months ago

jia1047 commented 5 months ago

I used the visualization code you previously provided to others, but when I used CC, the visualization result was blank and I don't know what's going on

image e78ff0f1f9442558e868c35ca0eee94

mellody11 commented 5 months ago

This may be caused by the representation range of colors. You should determine whether the representation range of colors is [0,255] or [0,1] in CC.

jia1047 commented 5 months ago

May I ask if I have used the visualization code you previously sent to someone else, which is writePly (os. path. coin (LOG-DIR, 'val_preds', name), [xyzrgb, preds, labels], ['x', 'y', 'z', 'red', 'green', 'blue', 'pred', 'label']), and its visualization results are the same as original_ply's visualization results. I have since modified the code and wrote two writePly codes. I want to obtain the visualization of real labels and predicted labels, but the visualization results of these two are the same. All are consistent, consistent with writePly (os. path. coin (LOG-DIR, 'val_preds', name), [xyzrgb, preds, labels], The visualization results obtained from ['x ',' y ',' z ',' red ',' green ',' blue ',' pred ',' label '] and original_ply are consistent. May I ask how your visualization code is specifically written? Can you obtain visualization of real labels and visualization of predicted labels 图片

mellody11 commented 5 months ago

It is foreseeable for the results to be consistent, because the rgb attribute is the same in your two ply files. As a result, if you directly visualize them, they are showing the same color.

if you want to visual the predicted result and gt like the following picture, you must map the predicted indices and gt indices to some specific colors and replace the rgb attribute then save as new ply files. 1