Filter non-curve AI detections from digitization and broaden synthetic plot typography#34
Merged
Merged
Conversation
Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/d53d58c6-04bf-4671-96d2-3e9b4807ab0e Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/d53d58c6-04bf-4671-96d2-3e9b4807ab0e Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
Copilot created this pull request from a session on behalf of
14NGiestas
May 15, 2026 12:40
View session
…ith tests Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/e2dd8d47-5e37-4283-abc1-8549be070e4a Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
…rd-compat comment Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/e2dd8d47-5e37-4283-abc1-8549be070e4a Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/1035d653-0ef9-41b2-90df-c3ef3af74b94 Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/ecdeb8e2-9037-44c5-9b2b-5148abd16c8e Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/5e70e96b-4d6f-43cb-b52f-4990f62ab98e Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
14NGiestas
requested changes
May 16, 2026
14NGiestas
left a comment
14NGiestas
left a comment
Owner
There was a problem hiding this comment.
When overlaying the segmentation results back to the original image, the scale. The point extraction after the curriculumn training run. Also it's not parallel and not interpreting the segments correctly: it probably found segments like axis and axis anchors, but not using them to extract the values correctly.
Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/f86f937b-059d-4393-965c-4c7abf84b704 Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
Agent-Logs-Url: https://github.com/14NGiestas/plot-digitizer/sessions/f86f937b-059d-4393-965c-4c7abf84b704 Co-authored-by: 14NGiestas <14851487+14NGiestas@users.noreply.github.com>
- Increased CURVE_MASK_PADDING_PIXELS to 8 to create a larger area around curves for improved segmentation training
- Added interactive paint_mask mode ('m') to annotate by painting over areas and automatically extracting points
- Updated test_digitizer.py to mock run_ai_segmentation to use cv_segmentation directly for test_generate_digitize_and_validate after strict AI flow enforcement
…zation - Switched default weights and curriculum models to yolo11n-seg.pt and yolo11s-seg.pt - Added a validation step in start.sh after each training phase that digitizes 5 images with --overlay - Used ThreadPoolExecutor in __init__.py to process multiple images in parallel during digitization
- Switched all curriculum stages and defaults from yolov8 to yolo11s-seg.pt - Using the Small (s) model variant consistently across all stages ensures we can safely transfer weights between curriculum stages without shape mismatches - Upgrading to Small instead of Nano improves segmentation recall on dense curves
- Fixed the 'mask shift' bug where ground-truth segmentation masks were slightly misaligned due to matplotlib's tight_layout() shifting axes after mask generation. The renderer now mathematically extracts the exact axes position after tight_layout() and renders the masks matching it perfectly. - Set rect: true in the curriculum runs to prevent the YOLO dataloader from scaling the rectangular synthetic plots into 640x640 squares. This preserves aspect ratio and prevents text stretching.
- The tests import the package naturally, so forcing PYTHONPATH=src breaks imports if the package is installed/exposed by Nix rather than strictly living uninstalled under src/
- Updated the smoke test script to look for '.csv' in 'digitized/csv/' and 'synthetic/csv/' instead of the old legacy locations ('digitized/' and 'synthetic/ground_truth/') which I removed in previous refactors
- Added a PLOT_DIGITIZER_SMOKE_TEST=1 env var override in digitize_workflow.py to cleanly fallback to OpenCV contours specifically during the CI smoke test since an untrained YOLOv11s model will intentionally fail the strict curve isolation check
- The mock in 'test_generate_digitize_and_validate' was failing because 'digitize_image' signature changed to include 'imgsz'. Updated the lambda to accept this argument.
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.
Uh oh!
There was an error while loading. Please reload this page.