Semestral work submitted in the course on visualization. The app allows to visualize activations and outputs from a neural network classifying MNIST digits and therefore to provide a better understanding of the model's behaviour.
The source data will be downloaded automatically and then visualized as points in the square window. For efficient exploration you can use filtering in the left panel or get the precise values by pointing at the tiles in the right panel.
- Install requirements:
pip install flask pillow numpy wget - Run app:
cd App
python run.py - Open browser (URL is shown in terminal)
Correctly classified input
Incorrectly classified input
