Issue when training model

ymli39 / DeepSEED-3D-ConvNets-for-Pulmonary-Nodule-Detection

DeepSEED: 3D Squeeze-and-Excitation Encoder-Decoder ConvNets for Pulmonary Nodule Detection

MIT License

109 stars 33 forks source link

Issue when training model #2

Closed bartmch closed 5 years ago

bartmch commented 5 years ago

Hi, I'm getting the following issue (after preprocessing the image). Would it be possible to look into this please? Also I couldn't find any ways in testing the model, before training it.

bartmch commented 5 years ago

On line 68: self.bboxes = np.concatenate(self.bboxes,axis = 0). The issue I'm facing is that self.bboxes = []. I'm running this in the training mode by setting the default value of --test default=0 so I can run the training procedure by just typing python train_detector_se.py

ymli39 commented 5 years ago

Are you using same data preprocessing code as I provided?

bartmch commented 5 years ago

Hi, yes - after running changing config_training.py and running prepare.py, I get all the preprocessed data in "luna_data" and "preprocess_result_path". I'm running into an empty bboxes error after I run train_detector_se.py in training mode. Thanks!

ymli39 commented 5 years ago

I did not encounter this problem before. I think you could set a debug point at dataloader line 57 to check whether it loads bounding boxes for each label?