The given repository builds a decent model for NSFW detection using NudeNetDataset with the help of Keras. Convolution Neural Networks are used for training the dataset. The model has been trained on a section of the dataset on local machines for the ease of deployment.

NudeNetDataset is used for training the model (a section of it). It comprises of three classes - nude (raw), sexy (can be passed further) and safe.

The Following models were used for extracting features (transfer learning). Also, image size was altered to check the dependencies.

• VGG16
• InceptionV3

VGG16 is used in the deployed model.

Make sure to have the requirements installed as provided in the requirements.txt.

pip install -r requirements.txt

Use the notebook provided in models/train.ipynb file in your preferred python execution setup.

The best model outputs will be automatically saved with the code provided in the file itself.

The code provided in the app.py file is used for testing the image and rendering the main page.

The model had been deployed using Flask web Framework. To run the model on localhost run:

python app.py 

Azure Deployment. (Decommisoned)

After spinning the server up online, just upload your image to get the predicted result.