NameError: name 'Program' is not defined

[zcy1065670505]

Hi ，Thank for your beautiful work on pointcloud semantic segmentation.When I try to run your code about SPG, I successfully in compiling and Sem3d data processing.But ouccr error when I try to train the network based on Sem3d.commands are as follows: CUDA_VISIBLE_DEVICES=0 python learning/main.py --dataset sema3d --SEMA3D_PATH /media/swjtu-rs/工作/ZCY/SEMANTIC3d --db_test_name testred --db_train_name trainval --epochs 500 --lr_steps '[350, 400, 450]' --test_nth_epoch 100 --model_config 'gru_10,f_8' --ptn_nfeat_stn 11 --nworkers 2 --pc_attrib xyzrgbelpsv --odir "results/sema3d/trainval_best"

Train dataset: 15 elements - Test dataset: 4 elements - Validation dataset: 0 elements Epoch 0/500 (results/sema3d/trainval_best): 0%| | 0/7 [00:01<?, ?it/s] Traceback (most recent call last): File "learning/main.py", line 459, in main() File "learning/main.py", line 329, in main acc, loss, oacc, avg_iou = train() File "learning/main.py", line 203, in train outputs = model.ecc(embeddings) File "/home/swjtu-rs/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in call result = self.forward(*input, *kwargs) File "/media/swjtu-rs/工作/ZCY/superpoint_graph-ssp-spg/learning/../learning/graphnet.py", line 97, in forward input = module(input) File "/home/swjtu-rs/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in call result = self.forward(input, **kwargs) File "/media/swjtu-rs/工作/ZCY/superpoint_graph-ssp-spg/learning/../learning/modules.py", line 176, in forward self._edge_mem_limit) File "/media/swjtu-rs/工作/ZCY/superpoint_graph-ssp-spg/learning/../learning/ecc/GraphConvModule.py", line 80, in forward ctx._degs_gpu.narrow(0, startd, numd)) File "/media/swjtu-rs/工作/ZCY/superpoint_graph-ssp-spg/learning/../learning/ecc/cuda_kernels.py", line 125, in conv_aggregate_fw function, stream = get_kernel_func('conv_aggregate_fw_kernel_v2', conv_aggregate_fw_kernel_v2(), get_dtype(src)) File "/media/swjtu-rs/工作/ZCY/superpoint_graph-ssp-spg/learning/../learning/ecc/cuda_kernels.py", line 38, in get_kernel_func prog = Program(ksrc, kname+dtype+'.cu')
NameError: name 'Program' is not defined

Could you tell me what's the problem? Thank you for your kind help!

[loicland]

Hi,

check line line 38 in /learning/ecc/cuda_kernels.py there are instructions on how to fix this issue.

[zcy1065670505]

Hi , I have solved this problem this morning. And successfully trained this Network by commonds: CUDA_VISIBLE_DEVICES=0 python learning/main.py --dataset sema3d --SEMA3D_PATH /media/swjtu-rs/工作/ZCY/SEMANTIC3d --db_test_name testred --db_train_name trainval --epochs 500 --lr_steps '[350, 400, 450]' --test_nth_epoch 100 --model_config 'gru_10,f_8' --ptn_nfeat_stn 11 --nworkers 2 --pc_attrib xyzrgbelpsv --odir "results/sema3d/trainval_best"

But I found the test test acc is too low : -> Test Loss: 10.8944 Test accuracy: 1.28% Test oAcc: 0.00% Test avgIoU: nan%

[loicland]

Please keep your issues focused.

This one is easy though, there are no ground truth for semantic3d for the reduced test set. You must write the label file with write_Semantic3 and submit it on the semantic3d.net servers.

[zcy1065670505]

I'm very appreciate for your repidly reply. there are several questions need your help:

No1. It means that if I use the commands below to train the Network based on Sem3D dataset. The test acc is based on test_full set or reduced test set? And Because of lacking ground truth(both of est_full set and reduced test set),so the acc is too low. CUDA_VISIBLE_DEVICES=0 python learning/main.py --dataset sema3d --SEMA3D_PATH /media/swjtu-rs/工作/ZCY/SEMANTIC3d --db_test_name testred --db_train_name trainval --epochs 500 --lr_steps '[350, 400, 450]' --test_nth_epoch 100 --model_config 'gru_10,f_8' --ptn_nfeat_stn 11 --nworkers 2 --pc_attrib xyzrgbelpsv --odir "results/sema3d/trainval_best"

No2. If I use commands below to train the network,the acc is validition acc? CUDA_VISIBLE_DEVICES=0 python learning/main.py --dataset sema3d --SEMA3D_PATH $SEMA3D_DIR --epochs 450 --lr_steps '[350, 400]' --test_nth_epoch 100 --model_config 'gru_10,f_8' --pc_attrib xyzrgbelpsv --ptn_nfeat_stn 11 --nworkers 2 --odir "results/sema3d/best".

No3. The acc in your paper is which acc?

Thank you!

[loicland]

1) --db_test_name testred -> it is the reduced test The ACC is not too low, it is nonsensical since there are no ground truth to compute it.

2) I think you should add --use_val_set 1 CUDA_VISIBLE_DEVICES=0 python learning/main.py --dataset sema3d --SEMA3D_PATH $SEMA3D_DIR --epochs 450 --use_val_set 1 --lr_steps '[350, 400]' --test_nth_epoch 100 --model_config 'gru_10,f_8' --pc_attrib xyzrgbelpsv --ptn_nfeat_stn 11 --nworkers 2 --odir "results/sema3d/best".

If the network is already trained, add --resume RESUME and make sure that the odir contains the trained model.

3) the results in the paper are from the semantic3d.net submission servers: http://www.semantic3d.net/view_results.php?chl=1

[zcy1065670505]

Thank you so much! It works.