Semantic segmentation tests its own dataset and reports errors

[bukeyan4318]

Hi, thanks for your nice work. When I tested semantic segmentation, I reported an error and couldn't understand it. I hope I can get your help. Thank you in advance. Traceback (most recent call last): File "test_semseg.py", line 202, in main(args) File "test_semseg.py", line 118, in main scene_data, scene_label, scene_smpw, scene_point_index = TEST_DATASET_WHOLE_SCENE[batch_idx] File "/shiyanshi2/user/sunyijun/code/Pointnet2-own/data_utils/S3DISDataLoader.py", line 164, in getitem data_room = data_room.reshape((-1, self.block_points, data_room.shape[1])) IndexError: tuple index out of range

[wyz5008]

hi,I am also testing my semantic segmentation dataset!The original pointcloud data has been collected,which file format is pcd. Could you tell how to mark labels?and change code to train it just like you?

[jiayaozi]

hello,same error,have you solved that?

[foysalahmed10]

@jiayaozi @bukeyan4318 @wyz5008 have you solved this problem?

[morankim]

Hi, I solved this by adding a 7th element in data array. maybe your dataset is composed of (x, y, z, r, g, b). transform this -> (x, y, z, r, g, b, label) where 0<= label <= num_class. but I just initialized lebel = 0 for all point cloud. and the result was fine. anyway, you need to give initial label index which can be set as random number btw 0~ num_class.