Experiments into Occlusion scoring heuristics for input prediction bias normalization #41
Conversation
This file contains 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
|
Kudos, SonarCloud Quality Gate passed!
|
Codecov Report
@@ Coverage Diff @@
## master #41 +/- ##
=======================================
Coverage 99.41% 99.41%
=======================================
Files 28 28
Lines 853 853
=======================================
Hits 848 848
Misses 5 5
Flags with carried forward coverage won't be shown. Click here to find out more. Continue to review full report at Codecov.
|
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Outdated suggestions cannot be applied.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.
We have found previously that there are deficiencies with the classifier-confidence-based saliency map generator when there perturbed image confidences that are imbalanced on the on the positive/negative scale when differenced from reference prediction results. This branch includes experiments into a "heuristic" (so named because I have not investigated a proof of generality) that aims add a "normalization" to counteract this bias.
Currently this is just experiments shown in-line in the image classification confidence saliency generation notebook. The PR is more for recordkeeping that an actual merge candidate.
The "winner" looks like heuristic # 3. This weights classification confidence difference vector based on the geometric mean of the >0 and <0 "classes". The intuition here is when there are mask regions that contribute to rare changes in prediction results, they must be relatively more important to that prediction.