Repository navigation
feat: add SAV tracks verifier resources server - #3507
Open
pulkitkumar95 wants to merge 1 commit into
Open
pulkitkumar95 wants to merge 1 commit into
pulkitkumar95 wants to merge 1 commit into
Conversation
SA-V tracks-schema verifier for point / box / text-referred tracking, used by combined SA-V + CapRL video GRPO training. This is the exact implementation the existing SA-V RL baselines trained against: per-object frame averaging, extras scale the reward down, configurable absence_score and iou_threshold, mask membership via pure-Python COCO RLE decoding. Includes config, README, example data, and tests.
This was referenced Oct 22, 2026
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds the
sav_tracksresources server: SA-V tracks-schema verifier for point / box / text-referred tracking (app.py, config, README, example data, tests).Why
Required for combined SA-V tracking + CapRL video GRPO. This is the exact verifier implementation the existing SA-V RL baselines trained against: per-object frame averaging, extras scale the reward down rather than zeroing it, configurable
absence_score/iou_threshold, mask membership via pure-Python COCO RLE decoding.Testing
Unit tests included under
resources_servers/sav_tracks/tests/. Validated end-to-end in multi-step async GRPO runs on 32 nodes (mixed SA-V + CapRL data, 120 training steps).