Name		Name	Last commit message	Last commit date
parent directory ..
Models/CNN		Models/CNN
README.md		README.md
__init__.py		__init__.py
load_data.py		load_data.py
load_images.py		load_images.py
predict.py		predict.py
split_dataset.py		split_dataset.py
train.py		train.py

README.md

Removing Non-Code and Noisy-Code Frames Using Deep Learning

The implementation is based on the source code from Ott et al.

Prepare Dataset

We use the web application to label frames as valid and invalid. The labels are stored in ../../webapp/labels/.
Then, we use the python script split_dataset.py to generate training and testing files in format of train.csv and test.csv.

Train

Please use the script train.py to train a model to predict images as valid or invalid

Prediction

Use the script predict.py to predict frames in our dataset