Inspired by the hotdog/not-hotdog classifier from HBO’s Silicon Valley S4E4.

Fixed feature extractor trained (with transfer learning) on only ~380 images for only 2 epochs.

git clone https://github.com/gianlucatruda/not-wors.git
cd not-wors
pip3 install -r requirements.txt

To run a demo, place any two pictures you like in /boeriebase/demo/unlabelled/ and run:

python3 inference.py models/ffe_1013.model demo

To run evaluation, simply run:

python3 inference.py models/ffe_1013.model val

• Gianluca Truda
• Zak Toyer
• Luke Neville
• Robbie Barnhoorn
• Luke Taylor

The code for this project is a significantly adapted version of the code presented in this tutorial, for which we thank Sasank Chilamkurthy.

The images in boeriebase/ were scraped from Google Images and Instagram. They are used for demonstration purposes only and do not fall under the same licence terms as the source code. The authors do not own the rights to the images.