Deep learning project that solves Where's Wally puzzles by finding the exact position of Wally in an image

HereIsWally is a Tensorflow project that includes a model for solving Where's Wally puzzles. It uses Faster RCNN Inception v2 model initially trained on COCO dataset and retrained for finding Wally using transfer learning with Tensorflow Object Detection API.

1. Install Python 3.7
2. Install Pipenv
3. Run pipenv install to install the dependencies
4. Run pipenv shell to activate the Pipenv environment

python find_wally_pretty.py images/1.jpg

or

python find_wally.py images/1.jpg

A window displaying a Wally puzzle outlining Wally should show up.

For instructions about how to retrain the model follow the instructions in this blog post.

• Tensorflow Object Detection API
• Training Images

• Gathering and Analyzing Hardware Performance Data During Deep Network Training

See LICENSE for details.