Neural Network for Filtering

The model.ipynb notebook contains a pre-trained ResNet18 neural network model. This model has been specifically trained to classify synthetic images based on their quality into two categories: "good" and "bad". The primary application of this model is to filter out undesirable images from a batch of synthetically generated content, thus enhancing the quality of datasets used in further applications.

Steps

1. Copy your synthetic images to the sort_images/ directory
2. Run the code blocks in the model.ipynb notebook, except these for training
3. The good images will be copied to results/good/ directory
4. Done!