Trained Model output resutl error about cylinder3d HOT 10 CLOSED

@YoushaaMurhij yes your correct, there was issue in the color encoding , thanks for point it out currently have resolved this issue now getting correct output

from cylinder3d.

Are you using RVIZ? Can you also show the GT?

from cylinder3d.

@YoushaaMurhij @xinge008 i am using pptk viewer to look into the annotations and predictions of semantic kitti data , below is the gt and right is predictions

from cylinder3d.

It feels like you have a color encoding problem .. no more, when converting the labels to colors.
Could u check that?

from cylinder3d.

how to get the visual results from raw network output?
the Cylinder3D network get [1, 480, 360, 32] input, and output is [1, 20, 480, 360, 32] torch tensor

res = model(train_pt_fea_ten, train_vox_ten, batch_size)
 
predicted_labels =  torch.argmax(res, dim=1) 

@abhigoku10 @YoushaaMurhij

from cylinder3d.

@muzi2045 i am actually saving the labels and then using semantic kitti code base to view the predicitions please find the reference which worked for me https://github.com/edwardzhou130/PolarSeg/blob/master/test_pretrain_SemanticKITTI.py

from cylinder3d.

thanks a lot, I'll try it.

from cylinder3d.

with the semantic-kitti api codebase, here is some result output:

why the road points are colored as black?
maybe the color map need to be change or something wrong when generate the points label file.
@abhigoku10

from cylinder3d.

@muzi2045 you have to use proper color map during saving and visualization

from cylinder3d.

@muzi2045 i am actually saving the labels and then using semantic kitti code base to view the predicitions please find the reference which worked for me https://github.com/edwardzhou130/PolarSeg/blob/master/test_pretrain_SemanticKITTI.py

when i use the same code as you, i got the error "ValueError: Input must be 1- or 2-d."---do you know what the problems ?

(pytorch) ziudarkin@ziudarkin-ubuntu:~/PolarSeg$ python train_SemanticKITTI.py
train_SemanticKITTI.py
Namespace(check_iter=4000, data_dir='data', grid_size=[480, 360, 32], model='polar', model_save_path='./SemKITTI_PolarSeg.pt', train_batch_size=2, val_batch_size=2)
  0%|                                                   | 0/136 [00:00<?, ?it/s]Traceback (most recent call last):
  File "train_SemanticKITTI.py", line 197, in <module>
    main(args)
  File "train_SemanticKITTI.py", line 129, in main
    iou = per_class_iu(sum(hist_list))
  File "train_SemanticKITTI.py", line 27, in per_class_iu
    return np.diag(hist) / (hist.sum(1) + hist.sum(0) - np.diag(hist))
  File "<__array_function__ internals>", line 6, in diag
  File "/home/ziudarkin/anaconda3/envs/pytorch/lib/python3.7/site-packages/numpy/lib/twodim_base.py", line 303, in diag
    raise ValueError("Input must be 1- or 2-d.")
ValueError: Input must be 1- or 2-d.
  0%|                                                   | 0/136 [00:01<?, ?it/s]

from cylinder3d.