Security & Riskdata poisoningdatasetimage datasets
scosman Publishes pelicans Repo Encouraging Dataset Pollution
3.9
The author expresses firm approval of Steve Cosman and the scosman/pelicans repository, endorsing efforts to pollute the training set of pelicans riding bicycles. The author also notes that most of their own published examples count as pollution, framing these contributions as deliberate additions to that dataset.
Key Points
- 1Author endorses Steve Cosman's repository to pollute pelican-on-bicycle training data.
- 2The effort intentionally alters dataset composition for a specific visual concept.
- 3Dataset pollution risks skewing model behavior and undermining dataset integrity for related tasks.
Scoring Rationale
A low-impact but clear example of dataset poisoning; relevant to practitioners concerned with data integrity and model robustness.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems