What

Are you a beginner ? You can take a jab at good first issues

Better documentation

• Document the models in our Robotoff models recap spreadsheet
• Increase the documentation of Robotoff itself
• Architecture schema for logos
• Write down the automatic application policy in plain English #969
• On all insight types, what do we extract, what do we use, what we don't use, what we validate automatically

Detections by technique, topic, field

Performance

• Improve performance of ML label/pictogram detection #371 (fellowship)
• Enable model retraining on validated detections #370 (fellowship)
• Improve automatic category detection by X% #369 (fellowship)
• Ingredients
• Refactor the OCR Category prediction as a datasource of the current category predictor

Support data quality

• Notification when anionique is found in an OCR
• Rich notification with support for moving to other projects