EE211_finalproject
├── caffe
│ ├── get_train_log.sh
│ ├── get_test_log.sh
│ └── models
│ │ │── bvlc_reference_caffenet
│ │ │ ├── train_val.prototxt
│ │ │ └── solver.prototxt
├── ilsvrc12_train_lmdb
│ ├── data.mdb
│ └── lock.mdb
├── ilsvrc12_val_lmdb
│ ├── data.mdb
│ └── lock.mdb
├── imagenet_mean.binaryproto
├── create_imagenet.sh
├── make_imagenet_mean.sh
├── train_test.sh
├── train.txt
└── val.txt
This project is Caffenet-based Automatic Prostate Tumor Classification from T2-MRI. The framework is caffe and we build our network based on BVLC reference caffenet and set our own solver file. We use PROSTATEx Challenge Data Sets for training and testing and random pick-in, random pick-out to augmentation the image. The final accuracy can get an average value of 0.7.
- Download data from our google drive which included in
data.md
. - Set up caffe environment and the instruction is included in
mac_caffe.md
. - Use our
train_val.prototxt
andsolver.prototxt
to replace the original files in models/bvlc_reference_caffenet. - Prepare data and use our script file to train model and get the result.
- Download data and set the fold name as train and val.
- Put train and val image fold and your caffe framework fold in your work space fold.
- Run
bash train_test.sh train
will create train.txt to save the train image path and its label in a txt file. - Run
bash train_test.sh val
will create val.txt in the same way. - Run
./create_imagenet.sh
will create train and val lmdb file which will help to increase the calculate efficiency. - Run
./make_imagenet_mean.sh
will create imagenet_mean.binaryproto which is useful in training.
- Make sure you have get train and val lmdb file, imagenet_mean.binaryproto file and you have build your caffe successfuly.
- Make sure you have successfully replced
train_val.prototxt
andsolver.prototxt
as the Dependencies part. - Get into your caffe fold and run
get_train_log.sh bvlc_reference_caffenet
to apply adjusted training file. If you want to test other network. You just need to rewrite your training file and change the model name to yours. - After training, you will find
train_result.txt
which saves the training log in caffe_train_log fold. - If you want to test the model, you can run
get_test_log.sh bvlc_reference_caffenet
which saves the predict log in caffe_net_log fold.
We save some logs of our training model in training_log fold as a sample.