{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreihbhhn2w7ykkhex4fjjd37rozk6m7ayebp3mjp45be3pe5zxd5a7q",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mizgeedgiav2"
  },
  "path": "/t/fixing-jagged-edges-in-clothes-parsing-segformer-rankseg/175079#post_1",
  "publishedAt": "2026-04-08T17:17:11.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "colab.research.google.com",
    "Google Colab",
    "GitHub - rankseg/rankseg: Boost segmentation model mIoU/Dice instantly WITHOUT retraining. A plug-and-play, training-free optimization module. Published in NeurIPS & JMLR. Compatible with SAM, DeepLab, SegFormer, and more. 🧩 · GitHub"
  ],
  "textContent": "**Hey everyone,**\n\nIf you’re using the `segformer_b2_clothes` model for downstream tasks like virtual try-on or outfit editing, you might have noticed that the standard `argmax` output can leave jagged, pixelated boundaries—especially around complex items like bags, skirts, and scarves.\n\nI recently experimented with replacing the `argmax` step with **RankSEG** , a metric-aware post-processing solver from NeurIPS 2025, and the visual difference is night and day. Best part? **No retraining required**.\n\n### The Hard Numbers (Tested on 500 images from the ATR dataset):\n\n  * **Global mDice** : Improved from 75.54% to 77.53%\n\n  * **Notable improvements in complex categories** :\n\n    * **Belt** : +10.49%\n\n    * **Skirt** : +3.59%\n\n    * **Bag** : +2.60%\n\n\n\n\nIf you’re curious or want to verify these results, I’ve packaged the entire pipeline into a self-contained Colab notebook. You can test it on your own images immediately:\n\ncolab.research.google.com\n\n### Google Colab\n\nAlso, here’s the GitHub - rankseg/rankseg: Boost segmentation model mIoU/Dice instantly WITHOUT retraining. A plug-and-play, training-free optimization module. Published in NeurIPS & JMLR. Compatible with SAM, DeepLab, SegFormer, and more. 🧩 · GitHub for the **RankSEG** implementation, so you can explore it further.\n\nI hope this helps anyone struggling with boundary artifacts! Feel free to try it out and let me know what you think.\n\nCheers,\nZhao Qingyang",
  "title": "Fixing jagged edges in clothes parsing (SegFormer + RankSEG)"
}