feat: add VGGT-Omega and LingBot-Map reconstruction backends - #37
Merged
Merged
Conversation
Add two feed-forward camera + depth reconstruction backends selectable via --mapping: - vggt_omega: whole-sequence forward pass; peak GPU memory grows with frame count. Gated FAIR checkpoint, noncommercial research license, pulled from GitHub via the vggt_omega extra. - lingbot_map: streaming frame-by-frame against a bounded KV cache; peak GPU memory stays flat. Public Apache-2.0 checkpoint, pulled from GitHub via the lingbot_map extra. Streaming and windowed modes, optional FlashInfer. Both predict per-frame intrinsics, so --refine-intrinsics-from-mapper defaults on for them. Adds point-cloud confidence/depth-edge filters, CLI options, model-zoo entries, README docs, and tests. Ignores locally cloned HF model repos (vggt/, etc.) so multi-GB checkpoints are never committed. Co-authored-by: Cursor <cursoragent@cursor.com>
Contributor
|
I would like to try to automatically generate an update for these tests if they fail. I'm thinking of a CICD task that will create a pull request into whichever branch an e2e test breaks in (for example, this one). So we can easily validate and update the golden tests if changes like this happen, I'll put the change here. |
Contributor
|
e2e golden regenerated: #38 |
Merged
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
HuguesSib
requested changes
Sep 23, 2026
Co-authored-by: Hugues Sibille <hugues.sibille@epfl.ch>
HuguesSib
approved these changes
Sep 24, 2026
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Add two feed-forward camera + depth reconstruction backends selectable via --mapping:
Both predict per-frame intrinsics, so --refine-intrinsics-from-mapper defaults on for them. Adds point-cloud confidence/depth-edge filters, CLI options, model-zoo entries, README docs, and tests. Ignores locally cloned HF model repos (vggt/, etc.) so multi-GB checkpoints are never committed.