Optimize watermark loading performance with PIL channel operations - #5
Open
aliciusschroeder wants to merge 1 commit into
Open
Conversation
Replace pixel-by-pixel iteration with PIL channel operations for watermark alpha channel modification. Benchmarks show 14.1x speedup (14.16ms → 1.00ms) with identical output quality. - Use Image.split() to separate RGBA channels - Apply transparency via alpha channel point operation - Maintain exact same transparency behavior - Remove code redundancy by refactoring loading into one single definition - No changes to memory footprint Performance validated on 150x108 test image across 1000 iterations.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This PR optimizes how watermarks are loaded by leveraging PIL's native channel operations instead of pixel-by-pixel iteration. This operation is critical as it runs for every single image during both training (477 × 3,111 = 1,483,947 patches) and testing (324 images).
Benchmark Results
Tested on 150x108 image over 1000 iterations:
Implementation
Validation
Impact
Given the training set of 1.48M patches and test set of 324 images, this optimization can play a significant role in optimizing the processing time during a complete training cycle (100 epochs). At 1,483,947 total patch operations, each millisecond saved in the watermark loading function translates to approximately 24.7 minutes reduction in total processing time during a complete training cycle.