[glenn-jocher]

Follow-up to #373 and #374 that resolves the open review questions on the comparison pages.

• EfficientDet tables (24 pages): the "CPU ONNX" column held the V100 GPU latencies from the google/automl README (e.g. D0 10.2 ms is "batch1 latency" on a single Tesla V100), and every row listed size 640. CPU cells are now "-", sizes are the native 512–1536 inputs from the EfficientDet paper (Table 1), bold marks the true best values, and the sentences and meta descriptions that relied on the CPU numbers are updated.
• YOLOv10 NMS-free wording (~60 pages): "pioneered" overclaimed, since NMS-free detectors such as DETR predate it. Pages now say YOLOv10 brought NMS-free training to YOLO models, matching the YOLOv10 paper (arXiv 2405.14458).
• Links CI: the two image-net.org links that time out on every Links run now point to the Ultralytics ImageNet dataset page.
• Other: fact fixes checked against the cited sources, an accurate TPU deployment note for EfficientDet, and consistent "arXiv" casing.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

WARNING ⚠️ this PR is very large, summary may not cover all changes.

🌟 Summary

Corrected model-comparison documentation by fixing EfficientDet benchmark data and narrowing claims about YOLOv10’s NMS-free contribution.

📊 Key Changes

• Updated EfficientDet tables across comparison pages with paper-reported input sizes, removed unsupported CPU latency values, corrected best-value highlighting, and revised related descriptions.
• Reworded YOLOv10 claims to say it introduced NMS-free training to YOLO models, rather than pioneering NMS-free detection broadly.
• Replaced two frequently timing-out ImageNet links with the Ultralytics ImageNet dataset page.
• Corrected EfficientDet’s TPU description and standardized “arXiv” casing across comparison pages.

🎯 Purpose & Impact

• Readers can compare EfficientDet using more accurately labeled benchmark data, and the documentation no longer attributes the broader origins of NMS-free detection to YOLOv10.

[UltralyticsAssistant]

👋 Hello @glenn-jocher, thank you for submitting a ultralytics/docs 🚀 PR! This automated message confirms your contribution was received, and an Ultralytics engineer will assist with the review. To ensure a seamless integration of your work, please review the following checklist:

• ✅ Define a Purpose: Clearly explain the purpose of your fix or feature in your PR description, and link to any relevant issues. Ensure your commit messages are clear, concise, and adhere to the project's conventions.
• ✅ Synchronize with Source: Confirm your PR is synchronized with the ultralytics/docs main branch. If it's behind, update it by clicking the 'Update branch' button or by running git pull and git merge main locally.
• ✅ Ensure CI Checks Pass: Verify all Ultralytics Continuous Integration (CI) checks are passing. If any checks fail, please address the issues.
• ✅ Update Documentation: Update the relevant documentation for any new or modified features.
• ✅ Add Tests: If applicable, include or update tests to cover your changes, and confirm that all tests are passing.
• ✅ Sign the CLA: Please ensure you have signed our Contributor License Agreement if this is your first Ultralytics PR; the CLA check will comment with signing instructions if needed.
• ✅ Minimize Changes: Limit your changes to the minimum necessary for your bug fix or feature addition. "It is not daily increase but daily decrease, hack away the unessential. The closer to the source, the less wastage there is." — Bruce Lee

For more guidance, please refer to our Contributing Guide. Don't hesitate to leave a comment if you have any questions. Thank you for contributing to Ultralytics! 🚀

[UltralyticsAssistant]

🔍 PR Review

_{Made with ❤️ by Ultralytics Actions}

Reviewed the comparison-page edits, including EfficientDet tables, NMS-free wording, and link replacements. One benchmark-resolution mismatch remains: the newly displayed native EfficientDet sizes do not match the input size used for the retained T4 latency measurements.

💬 Posted 1 inline comment

• 💡 MEDIUM docs/en/compare/damo-yolo-vs-efficientdet.md:61 The row now reports EfficientDet-d0 at 512 pixels, but the retained 3.92 ms T4 TensorRT latency comes from a benchmark run at 640×640; the benchmark report specifies 640×640 inputs, and the repository benchmark data records the EfficientDet entries at size 640. This leaves the displayed input size and latency inconsistent (the same issue affects the other rows whose native sizes differ from 640). Please remeasure T4 latency at the displayed native sizes or distinguish the benchmark input size f…

[UltralyticsAssistant]

🎉 Thanks, @glenn-jocher, for the merged PR!

“Well done is better than well said.” — Benjamin Franklin

These updates make the model comparisons and YOLOv10 claims more precise and better grounded in the cited sources.