Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
55 lines (36 loc) · 3.85 KB

CONTRIBUTING.md

File metadata and controls

55 lines (36 loc) · 3.85 KB
Raw

Maintenance and Contributing

English | 简体中文

Contributions of oriented object detection papers, author-maintained code, and bilingual reading notes are welcome. The scope covers 2D oriented bounding box detection and directly related representations, losses, features, supervision, and deployment methods.

Adding a Paper

  1. Open the original source and verify the full title, publication status, and year. Prefer official conference websites and proceedings; qualify acceptance claims when only an author's statement is available. Submission does not imply acceptance.
  2. Check for duplicates using the arXiv ID (without the version suffix), DOI, title, and authors. Keep one main entry for a preprint and its conference version; list independent papers introducing new methods separately.
  3. Add a row under the paper's primary topic in README.md, in reverse chronological order within that topic.
  4. Add questions to investigate, reading and reproduction status, code availability, and the verification date to TRACKING.md. New entries start as To read and Unverified.
  5. Record additions or changes in CHANGELOG.md, and update the counts and latest verification date on the homepage.
| 2026 | arXiv preprint | Full Paper Title | [paper](paper-url) · [code](author-code-url) |

Use code: unverified when official code cannot be confirmed. Clearly label a baseline implementation or partial release. Do not label third-party implementations as official code.

Adding a Note

Copy templates/paper-note.md to:

docs/<VenueYear>/oriented_object_detection/<short-name>.md

For example, docs/CVPR2025/oriented_object_detection/gaucho.md. Papers without a formal publication can use docs/arXiv2026/; update the path and all links after formal acceptance.

Distinguish three stages in each note:

  • Abstract screening: original links, the problem, an abstract summary, and questions for a close reading. Do not add experimental numbers that have not been verified.
  • Close reading: explain the method, cite the relevant equations or figures, and document experimental settings, evidence, and limitations.
  • Reproduction: record the environment, code commit, configuration, commands, raw results, and conclusions.

Link the note from both the README's Links column and the reading tracker. Distinguish the authors' claims, our interpretation, and our measured results. Keep application ideas as questions to validate.

Maintaining Both Languages

  • Use English in the default .md file and the same filename with .zh-CN.md for Simplified Chinese. Keep both versions in sync when adding or changing content.
  • Keep the language switch at the top of each page and link to the matching language for internal navigation. When adding a note, update both language versions of the paper list and reading tracker.

Update Checklist

  • Relative links and section anchors work, table column counts are consistent, and paper counts are correct.
  • Update existing entries when acceptance status, arXiv versions, or code availability change; do not add duplicates.
  • Update a paper's verification date only after checking it. Do not set every date to today to suggest freshness.
  • Do not rank results directly unless they use the same dataset version, split, backbone, tiling, input, TTA, and evaluation settings.
  • For latency measurements, record the GPU, batch size, precision, warm-up, sample count, synchronization method, and timing boundaries.
  • Preserve attribution for original papers, images, external code, and data. Prefer storing links and original notes in this repository.

Maintenance is currently manual. Record the scope, sources, and changes for each search; a check may be logged even when no papers are added. Scheduled collection has not been configured.

Back to the paper list