ProjectSidewalk / ProjectSidewalk/RampNet

Document the YOLOv11-pano training instability as a reported caveat

Open
#72 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

documentation enhancement
Dominant language
Python
Stars
7
Forks
1
Avg merge
4d 11h
Merged PRs (30d)
7

Description

Parent: #51

If the stabilized rerun (the v11-pano stabilization issue) confirms YOLO11 is fragile at
pano resolution while YOLO26 is stable, that is a legitimate, reportable finding
capture it rather than bury it:

"YOLO11 exhibited training instability at 1280px pano resolution with the small physical
batch the input size permits (validation mAP collapsed to ~0 for several epochs despite
healthy, still-decreasing training loss and no NaN/AMP error); YOLO26 trained stably
under the same budget."

The supporting evidence is already committed at scripts/model_comparison/yolo_baseline/
(per-epoch results.csv showing the peak→zero curve, resolved args.yaml, provenance).

Reporting this pre-empts a reviewer independently discovering the instability, and makes
the choice to report the YOLO26 / stabilized-v11 numbers transparent rather than looking
like selective omission.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reviewing the stabilized rerun referenced by the parent issue, then inspect scripts/model_comparison/yolo_baseline/, especially the per-epoch results.csv, resolved args.yaml, and provenance. Document the YOLO11 instability and the YOLO26 comparison once the rerun confirms the behavior, preserving the reported training and validation details.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
documentation, machine-learning
Issue type
Documentation
Difficulty
3/5
Estimated time
1-2 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.