Skip to content

Implement image-6 text removal and align generation defaults#5

Open
jaejung-dev wants to merge 4 commits intolica-world:mainfrom
jaejung-dev:text-removal-implementation
Open

Implement image-6 text removal and align generation defaults#5
jaejung-dev wants to merge 4 commits intolica-world:mainfrom
jaejung-dev:text-removal-implementation

Conversation

@jaejung-dev
Copy link
Copy Markdown
Contributor

Summary

  • Implement image-6 (Text Removal & Background Inpainting) pipeline end-to-end in GDB, including data loading, model I/O handling, and evaluation integration.
  • Add ReMOVE metric support (src/gdb/metrics/remove_metric.py) and PSNR metric wiring (src/gdb/metrics/core.py, src/gdb/metrics/__init__.py), and classify image-6 under typography while preserving benchmark ID compatibility.
  • Align generation behavior with current policy by removing forced masked composition in FLUX.2 local generation, making typography-8 mask-free at model-input time, and updating diffusion defaults/docs from flux.2-klein-4b to flux.2-klein-9b.

Test plan

  • python scripts/run_benchmarks.py --list
  • Regenerate impacted FLUX outputs on GPU 1 and replace prior artifacts for layout-8 and typography-7.
  • Regenerate typography-8 outputs after removing mask metadata and verify summary/log updates.
  • Validate that typography-8 ModelInput no longer includes mask metadata.

Made with Cursor

…r defaults.

This adds the text removal task pipeline and ReMOVE/PSNR metric wiring, removes forced masked composition behavior, updates typography-8 to run without mask metadata, and switches default diffusion runs/docs to FLUX.2 klein 9B for consistent baseline behavior.

Made-with: Cursor
@jaejung-dev jaejung-dev requested a review from purvanshi as a code owner April 20, 2026 12:50
This updates the smoke assertion to the current 40 benchmark registry size and applies Ruff-compliant import sorting in remove_metric so lint passes.

Made-with: Cursor
Keep a compatibility shim in tasks/image.py and align README/HELM metadata so image-6 remains discoverable as a typography-domain benchmark with consistent benchmark counts.

Made-with: Cursor
Apply Ruff-compatible import grouping in typography.py only so the PR passes lint without changing runtime behavior.

Made-with: Cursor
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant