Folder objectrecognition:
- neat - neat-python dependencies
- object - code to process images and train using NEAT
TODO:
Use more samplesUse train and test setSave best fit individual- Find a better way to document all results
- Implement a better visualization to use for all results
- Find out which samples (pos/neg) are the most problematic
- Decrease number of inputs - 50?
- Use different convolutional filters
- Use normal (wide angle) images
- Use soft max layer?
- Change path checkpoints and add .gitignore
TODO report:
- Write introduction to NEAT
- ...