fix: ocr visualization and add ocr recognition metrics

[samiuc]

Visualization Fixes

• Updated side-by-side canvas to compare ground truth (GT) vs. predictions, with per-word overlays.
• For mismatched true positives, the overlay shows the label as: pred_text
• Included a stitched legend and summary statistics (TP/FP/FN counts, Precision, Recall, and F1 score).

Recognition Metrics

• New metrics added (previously we only had the detection metrics, now we have detection + recognition metrics):
  • Word accuracy (case-sensitive and case-insensitive).
  • Character accuracy (case-sensitive and insensitive), using union-based edit distance.
• Edit distance is calculated using standard Levenshtein:
  • String trimming or character normalization is applied.
  • Case-insensitive metrics are computed using uppercase versions of the strings.
• Aggregation is union-based by default (configurable).
• Recognition mismatches:
  • Explicitly flagged as is_true_positive=False in metadata.
  • Shown as incorrect in the visualization.
• Ignore zone:
  • Added Optional IoU-based (HWR) - currently not used but can be enabled.

Word Merging and Weighting Fixes

• Added word_weight to correctly assign credit for merged GT words in recognition metrics.
• Merged words now retain both text and orig values:
  • GT orig uses the ground truth text.

Ignore-Zone Filtering

• Default filter: Removes words with less than 10% axis overlap relative to word size.
• HWR filter: Removes GT or predicted zones with:
  • IoU ≥ 0.3, or
  • ≥ 95% overlap on both axes (near-contained).
• Filter type is selectable via ignore_zone_filter_type ("default" or "hwr").

Aggregation and Reporting

• Detection metrics (Precision, Recall, F1) are calculated at the dataset level.
• Recognition aggregation mode is configurable:
  • Default is "union", with "mean" also available.
• Per-document evaluation metadata includes:
  • TP pairs with edit distances,
  • FPs and FNs

[github-actions]

❌ DCO Check Failed

Hi @samiuc, your pull request has failed the Developer Certificate of Origin (DCO) check.

This repository supports remediation commits, so you can fix this without rewriting history — but you must follow the required message format.

---

🛠 Quick Fix: Add a remediation commit

Run this command:

git commit --allow-empty -s -m "DCO Remediation Commit for samiullahchattha <Sami.Ullah1@ibm.com>

I, samiullahchattha <Sami.Ullah1@ibm.com>, hereby add my Signed-off-by to this commit: 794d99fc12d4fe0dcc6debc77ee05fcbd07730e4
I, samiullahchattha <Sami.Ullah1@ibm.com>, hereby add my Signed-off-by to this commit: 20b89544994c0c12c34721b32eb38e230b690f36"
git push

---

🔧 Advanced: Sign off each commit directly

For the latest commit:

git commit --amend --signoff
git push --force-with-lease

For multiple commits:

git rebase --signoff origin/main
git push --force-with-lease

More info: DCO check report

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[divekarsc]

LGTM!

[PeterStaar-IBM]

lgtm!