determine how to assess model accuracy
epierotti3 opened this issue · comments
Elizabeth Pierotti commented
Overview
We need to see how accurate the current model is (without augmentation) before we implement augmentation in order to see what improvements we can make. And will be useful for checking other model improvements (e.g., if we ever leave yolo v5). Some research needs to be done ahead of time to determine how to do this.
Action Items
- Get the docker container working on Alex's computer (may need to get help from more experienced contributors)
- Figure out how to run the model as it stands
- Determine best way to assess model accuracy:
- Look over methods used by DeepLearning.AI
- Model comparison methods in Python
- Will be great if it has a visualization aspect (so easy to share results with others)
- Implement model accuracy test (this will be its own issue)
Resources/Instructions
DeepLearning.AI
Google Scholar (Christian might know of some articles that talk about this)