Description
- Use GCV to perform both content moderation and annotation in a single request (if possible).
- If the dataset contains a "LIKELY" or "VERY_LIKELY" rating for inappropriate content for at least one of the images, the job should NOT be launched and funds NOT refunded.
- If the dataset doesn’t contain a "LIKELY" or "VERY_LIKELY" rating, but contains "POSSIBLE" ratings, send a notification to slack.
- In case of getting the dataset fully annotated, do not launch the escrow on chain and create the results file.
- Datasets from whitelisted users can skip the content moderation step but should still undergo annotation.
Implementation details
To be defined by developer
Description
Implementation details
To be defined by developer