Suggestive Annotation:
- Download the gland segmentation challenge dataset and extract it into folder "Warwick QU Dataset (Released 2016_07_08)".
- To change all images to same dimensions run, $python reshape.py It generates the new images in reshaped_warwick directory. Move the images you don't want to start training with, into images_train_eval and also the corresponding annotations into segments_train_eval.
- Now to generate the csv files, run $python gen_csv.py
- Now, to run 4 sessions of training over 4 bootstrapped datasets, run $python train.py
- To get the list of images to add for active learning, run $python eval.py # Make sure the eval_data flag in eval.py is train_eval
- After this, to get the list of images to add to the training set (move from train_eval to train folder) run $python active_selection.py
Again goto step 3, till satifactory results are not obtained.