Add CORE postprocessor code: Confidence + Orthogonal Residual Evidence#312
Open
JinMoYang wants to merge 1 commit into
Open
Add CORE postprocessor code: Confidence + Orthogonal Residual Evidence#312JinMoYang wants to merge 1 commit into
JinMoYang wants to merge 1 commit into
Conversation
Post-hoc OOD detection method that decomposes penultimate features into components parallel and orthogonal to classifier weight directions. Combines energy score with residual direction consistency via z-score normalization. Uses only ID training data (no OOD leakage). - Add COREPostprocessor implementation - Add core.yml config - Register CORE in postprocessor registry and __init__ - Add get_fc() to RegNet_Y_16GF for classifier weight extraction
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 CORE postprocessor
Adds CORE (Confidence + Orthogonal Residual Evidence), a post-hoc OOD detector that decomposes penultimate features into components parallel and orthogonal to classifier weight directions and combines a parallel (energy) and orthogonal
(class-mean cosine) score.
Paper: https://arxiv.org/abs/2603.18290
Leaderboard PR: zjysteven#6
Changes
openood/postprocessors/core_postprocessor.pyconfigs/postprocessors/core.ymlopenood/postprocessors/__init__.py,openood/evaluation_api/postprocessor.py— registration + hookup under the'core'keyopenood/networks/regnet_y_16gf.py— addsget_fc()for run on RegNetUsage
```python
from openood.evaluation_api import Evaluator
evaluator = Evaluator(net, id_name='imagenet', postprocessor_name='core', ...)
```
Results
Leaderboard entries (CIFAR-100, ImageNet-200 ×2, ImageNet-1k ×5) submitted as a separate PR. Notable: ImageNet-1k OOD Far-OOD AUROC = 98.00, Near-OOD AUROC = 90/33 on RegNet_Y_16GF.