Wildbook: Software and Machine Learning Engineers Join the Fight...
summary
Wildbook is an open source data platform that pairs a Java J2EE data server and Python ML pipeline with species‑specific ID models (triplet‑loss embeddings, trailing‑edge matching, detection and segmentation). Engineers adapt the groth star pattern algorithm from astronomy, run aerial surveys and rallies, and ingest social videos; the system now tracks 131,000 animals across 6 million photos with ~700,000 sightings and serves about 900 researchers. Deployed services automate detection, ID and iterative monitoring and have produced many validated new individual records.
AI-written summary · about
chapters (18)
- 0:00Intro
- 2:18Jason Holmberg
- 12:11Challenge: Identify them in pictures!
- 20:25Start small: SeadragonSearch
- 22:32We Are "Wild Me"
- 23:52Iterate to fight extinction
- 27:06Monitor animals continuously
- 28:39Census with rallies
- 29:59Survey for Population Assessment
- 31:30Animal detection
- 32:22Detection classification
- 33:25Coarse background segmentation
- 33:59Orientation
- 34:47ID Algorithms for Matching
- 35:32Photobombs and scenery matching
- 36:40NEW: Deep Learning for Matching
- 37:49Scale It: Graph ID without human decisions
- 38:52Intelligent Agents in Partnership
related talks
59mFaces, Flukes, Fins, and Flanks: How Multispecies Re-ID Models are Transforming Our Approach to AI..
11mUsing AI to track wildlife patterns
15mThe Internet of Wild Things: Acoustics, AI and Conservation
05mBringing AI into the Wild
40mWho's singing? Automatic bird sound recognition with machine learning
48mSetting the stage for AI for biodiversity
24mSven Kreiss, Lead Data Scientist, Wildcard
28mConservation & Tech