Pathology_Analysis

Pathology_Analysis

files

./read: take network.bin and generate data.csv
./Preprocess.ipynb: Generate splits. Use all 1430 pos and sample 2860 neg to generate 3300 (train) + 990 (test) dataset split.
./resnet/go.sh: train from scratch and show the result
./resnet/go_transfer.sh: transfer from pre-trained model on cifar10 and train the model for target pathology data
./vis.ipynb: visualize the result (from logs under ./resnet/checkpoint)

bash usage

(under the resnet folder) (dependency: pytorch)
bash go.sh
bash go_transfer.sh

command line guidance

Please try python main.py -h

PyTorch cifar10 for pathology

-h: show this help message and exit
--lr: learning rate
--lr_step: lr_step
--epoch: epoch
--r: resume from checkpoint
--sp: splits
--img_dir: img dir
--id: model id
--transfer: transfer to new task with pretrained
--pretrained: the pretrained model