Support tiling#6955
Open
A-Artemis wants to merge 46 commits into
Open
Conversation
…ms (skipping yolo/ultralytics)
github-actions
Bot
added
Geti Backend
⏱️ Backend import time —
|
| Accelerator | Cumulative (s) | Self (s) |
|---|---|---|
cpu |
4.901 | 0.002 |
xpu |
3.777 | 0.002 |
cuda |
4.945 | 0.002 |
📊 Test coverage report
|
🐳 Docker image sizes
|
kprokofi
reviewed
Jul 1, 2026
kprokofi
reviewed
Jul 1, 2026
…ing_extensions into aurelien/support-tiling
…h detection and instance segmentation.
leoll2
reviewed
Jul 14, 2026
…into aurelien/support-tiling
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 67 out of 67 changed files in this pull request and generated 5 comments.
Comments suppressed due to low confidence (1)
library/src/getitune/data/dataset/tile.py:1
- Calling
patch_instance_mask_tiler()at import time introduces global side effects (mutating Datumaro'sTilerRegistry) whenever this module is imported, even in non-tiling runs/tests. Consider moving the patch behind an explicit initialization point (e.g., insideTileDatasetFactory.create(...)or just before creating tiling transforms) while keeping idempotency—this reduces surprising behavior and makes the patch easier to reason about and test.
A-Artemis
commented
Jul 17, 2026
…into aurelien/support-tiling # Conflicts: # application/backend/app/execution/training/getitune_trainer.py # library/src/getitune/backend/lightning/models/detection/base.py # library/src/getitune/backend/lightning/models/instance_segmentation/base.py
kprokofi
reviewed
Jul 17, 2026
| # Copyright (C) 2026 Intel Corporation | ||
| # SPDX-License-Identifier: Apache-2.0 | ||
|
|
||
| """Runtime patches for Datumaro's experimental tiling. |
Contributor
There was a problem hiding this comment.
This should be addressed on the Datumaro side. Can you create a PR there?
Contributor
Author
There was a problem hiding this comment.
| compute_kwargs = {"best_confidence_threshold": best_confidence_threshold} | ||
| return super()._compute_metrics(metric, **compute_kwargs) | ||
|
|
||
| def forward_tiles(self, inputs: TileBatchData) -> PredictionBatch: |
Contributor
There was a problem hiding this comment.
It should be done via ModelAPI Tiler: https://github.com/open-edge-platform/model_api/blob/master/model_api/src/model_api/tilers/instance_segmentation.py
Why do we need custom forward for OV models?
Contributor
Author
There was a problem hiding this comment.
leoll2
reviewed
Jul 17, 2026
Contributor
|
@A-Artemis can you add tiling recipes for the new edge crafter model? In the same way as for DEIMV1/V2. They share similar code base |
…into aurelien/support-tiling # Conflicts: # application/backend/app/core/logging/utils.py
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.
Summary
Added
TilingDatasetFactoryfor wrapping the dataset subsets if tiling is enabled.Created tiling config yamls for algorithms which were missing it.
Added basic tiling support across data module and training pipeline.
Checklist