search

lppllppl920 / EndoscopyDepthEstimation-Pytorch

Official Repo for the paper "Dense Depth Estimation in Monocular Endoscopy with Self-supervised Learning Methods" (TMI)
GNU General Public License v3.0
124 stars 27 forks source link

Verification set #3

Closed resindraburiza closed 4 years ago

resindraburiza commented 5 years ago

Hello, first of all, thank you for sharing the code to the public.

I tried your code and, even though I had to make some modification, I can run the training phase without any problem. But, there is no validation set to validate the training result. Is it possible for you to provide a validation set too? Or, did I do something wrong? Can the validation also be done using the data in this repository?

Thank you so much.

resindraburiza commented 5 years ago

Also, I have an additional question.

When I run teacher_training.py, I got this result:

Epoch 4, lr [0.00011080000000000001]: 100%|█| 40/40 [00:03<00:00,  8.06it/s, loss=20.90089 15.47590, loss_depth_consistency=8.06786 9.90660, loss_scale_std=6.74215 2.22112, loss_sparse_flow=6.09088 3.34819]('Failure sequences: ', {'example_training_data_root/bag_1/_start_004259_end_004629_stride_25_segment_13/': 1})
Number of failure sequences: 1

what does it mean? Does it mean the training was not successfully done? Why is it so? Thank you very much.

cyh4520 commented 4 years ago

Also, I have an additional question.

When I run teacher_training.py, I got this result:

Epoch 4, lr [0.00011080000000000001]: 100%|█| 40/40 [00:03<00:00,  8.06it/s, loss=20.90089 15.47590, loss_depth_consistency=8.06786 9.90660, loss_scale_std=6.74215 2.22112, loss_sparse_flow=6.09088 3.34819]('Failure sequences: ', {'example_training_data_root/bag_1/_start_004259_end_004629_stride_25_segment_13/': 1})
Number of failure sequences: 1

what does it mean? Does it mean the training was not successfully done? Why is it so? Thank you very much.

hi i have some problem on the training phase. Can u help me. when I run the training code as the author provides. It came to be

File "/share/chyh/workspace/EndoscopyDepthEstimation-Pytorch/dataset.py", line 353, in getitem pair_extrinsic_matrices = [self.extrinsics_per_seq[folder][pos], KeyError: 'example_training_data_root/bag_1/_start_004259_end_004629_stride_25_segment_13'

and my training script is python student_training.py --id_range 1 11 --input_downsampling 4.0 --network_downsampling 64 --adjacent_range 5 30 --torchsummary_input_size 256 320 --batch_size 2 --num_workers 8 --validation_interval 1 --display_interval 50 --dcl_weight 0.1 --sfl_weight 20.0 --max_lr 1.0e-3 --min_lr 1.0e-4 --inlier_percentage 0.99 --testing_patient_id 4 --load_intermediate_data --use_hsv_colorspace --number_epoch 2 --num_iter 100 --architecture_summary --training_result_root "result" --training_data_root example_training_data_root/

thank you very much

lppllppl920 commented 4 years ago

I'm sorry for not responding to the issues in time. I have now fixed the issues related to running the training script. In terms of the validation set, because the sinus dataset is proprietary, we cannot share more data. I'm sorry for the inconvenience.