WebVision is an exploratory project which wants to find ways to learn from noisy data.
The process pipeline is as follows:
- Filter data
- Train baseline model
- Extract features and clustering
- Design curriculim learning strategy
- Train models from easy to hard
Please check out this site for dataset details.
