Codex/readme table library headers#66
Conversation
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b5b1c5dafd
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| for key, value in load_results_dir(VIDEO_RESULTS).items() | ||
| if value["media"] == "video" and value["library"] != "albumentations_mit" |
There was a problem hiding this comment.
When the generated paper CSV has no video rows (as in the checked-in docs/paper_data/all_results.csv, whose video16f_* CSVs are header-only), this fallback builds the README video table from output_videos. Those JSON files are standalone micro runs from different environments (AlbumentationsX on macOS arm/50 videos, Kornia/TorchVision on an RTX 4090/200 videos), so the resulting speedup and bold winner cells compare hardware and workloads rather than the selected benchmark regimes above. This should either omit the video table until same-regime published rows exist or require comparable video artifacts.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Pull request overview
Updates the README’s “Benchmark Results” presentation to use scenario-specific columns with library headers (and adds a generated video table), with supporting changes in the paper figure/table generator script.
Changes:
- Adjusts
scripts/paper/generate_figures_and_insights.pyto generate “Result Tables” with per-column library headers and to optionally populate the video table fromoutput_videos/viatools.compare. - Restructures
README.mdto include the new “Benchmark Results” TOC and updated generated tables/wording (including video table content). - Renames the winner-counts appendix figure label to “Figure 5” in both generator and README output.
Reviewed changes
Copilot reviewed 2 out of 7 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
| scripts/paper/generate_figures_and_insights.py | Changes README table generation to use fixed (regime, library) columns and adds a fallback that renders video tables from output_videos/ via tools.compare. |
| README.md | Replaces the previous benchmark-table sections with the new “Benchmark Results” structure and updated generated tables. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| @@ -512,33 +317,7 @@ For AlbumentationsX, each clip is a NumPy array `(T, H, W, C)`. The built-in spe | |||
| | WaterRefraction | **22 ± 0** | - | - | N/A | | |||
| | ZoomBlur | **4 ± 0** | - | - | N/A | | |||
|
|
|||
| sys.path.insert(0, str(ROOT)) | ||
| from tools.compare import format_comparison_table, load_results_dir | ||
|
|
| active_columns = [ | ||
| (regime, library) | ||
| for regime, library in columns | ||
| if ((subset["regime"] == regime) & (subset["library"] == library)).any() | ||
| ] | ||
| headers = [ | ||
| "Transform", | ||
| *[ | ||
| str(subset.loc[subset["regime"] == regime, "regime_label"].iloc[0]) | ||
| for regime in regimes | ||
| if (subset["regime"] == regime).any() | ||
| f"{LIBRARY_DISPLAY.get(library, library)}<br>{subset.loc[subset['regime'] == regime, 'regime_label'].iloc[0]}" | ||
| for regime, library in active_columns | ||
| ], | ||
| ] | ||
| active_regimes = [regime for regime in regimes if (subset["regime"] == regime).any()] | ||
| blocks.append("| " + " | ".join(headers) + " |") | ||
| blocks.append("| " + " | ".join(["---", *["---:" for _ in active_regimes]]) + " |") | ||
| blocks.append("| " + " | ".join(["---", *["---:" for _ in active_columns]]) + " |") | ||
| for transform in sorted(subset["display_transform"].unique()): |
No description provided.