Back to the episode map

Research Note

Point Tracking Deployment Review Record

Freeze the exact repository revision, checkpoint hash, dependency lock, model mode, preprocessing, configuration, and evaluation record. Record code, model, dataset, medi

Aug 4, 20262 min readBy Dalton Anderson

Point Tracking Deployment Review Record

Artifact boundary

Freeze the exact repository revision, checkpoint hash, dependency lock, model mode, preprocessing, configuration, and evaluation record. Record code, model, dataset, media, and dependency licenses separately.

The official CoTracker repository says most of the project uses a CC BY-NC license, with some components under other terms. This is a commercial-use and provenance gate, not a footnote.

Pipeline review

LayerEvidence
CaptureCamera, source, consent, rights, time, location
IngestionCodec, validation, malware boundary, dropped frames
PreprocessingDecode, resize, crop, color, frame rate, normalization
InferenceCheckpoint, mode, queries, windows, hardware, batching
OutputCoordinate system, visibility, confidence, schema, consumers
StorageRaw video, derived tracks, retention, encryption, access
MonitoringLatency, memory, queue, drift, sampled failures, incidents
ControlHuman review, abstention, stop condition, rollback, deletion

Runtime

Measure end-to-end capture-to-consumer latency, not only model time. Test warm and cold starts, different query counts, resolution, long streams, queue growth, device contention, corrupted input, dropped frames, and hardware loss.

Online mode is not automatically real-time. Offline mode may use future frames and may be unsuitable for a causal decision. The selected mode must match the workflow.

Silent failure

A crashed job is visible. A plausible but wrong trajectory can pass downstream unnoticed. Monitoring should sample rendered tracks, compare with human-reviewed clips, detect impossible jumps or stale coordinates, and measure the consequence in the consuming workflow.

Release stages

Start with offline replay, then shadow mode, then a bounded assisted workflow, then a narrow production release. Each stage requires an owner, threshold, human override, incident path, and exercised rollback.

Governance boundary

The deployment record must state purpose, affected people, access, data use, prohibited uses, monitoring, complaints, and reapproval triggers. A new camera, population, purpose, market, or decision consequence requires renewed review.

This framework is not authorization for surveillance, autonomous control, medical use, employment monitoring, or another high-stakes application.

Sources

Follow the evidence.

  1. youtu.be: BNTcjZ0Ym38youtu.be
  2. ai.meta.com: sam2ai.meta.com
  3. proceedings.neurips.cc: 58168e8a92994655d6da3939e7cc0918 Abstract Datasets and Benchmarksproceedings.neurips.cc
  4. arxiv.org: 2410arxiv.org
  5. open.spotify.com: 26JgnnwjvK5vYIdRofV8ntopen.spotify.com
  6. github.com: co trackergithub.com
  7. cotracker3.github.iocotracker3.github.io
  8. NIST AI Risk Management Frameworknist.gov
  9. vggsfm.github.iovggsfm.github.io
  10. daltonanderson.ghost.io: metas cotracker 3 a leap in ai object trackingdaltonanderson.ghost.io
  11. arxiv.org: 1803arxiv.org
  12. ecva.net: 3526 ECCV 2020 paperecva.net
  13. github.com: tapnetgithub.com
  14. raw.githubusercontent.com: LICENSEraw.githubusercontent.com
  15. tapvid.github.iotapvid.github.io
  16. NIST Privacy Frameworknist.gov
  17. arxiv.org: 1504arxiv.org
  18. openaccess.thecvf.com: Karaev CoTracker3 Simpler and Better Point Tracking by Pseudo Labelling Real Videos ICCV 2025 paperopenaccess.thecvf.com
Point Tracking Deployment Review Record