Itbanque/fashion_segformer

Fashion Segmentation (Combined: DeepFashion2 + Bikini)

A multi-class semantic segmentation dataset that merges:

• DeepFashion2-style labels from https://huggingface.co/datasets/sayeed99/fashion_segmentation
• Bikini segmentation from https://universe.roboflow.com/bikini-segmentor/bikini-detector-v24

An extra class bikini is added on top of the 46 DeepFashion2 categories. Suited for models like SegFormer, U-Net, etc.

---

Contents

• Images: RGB fashion photos.
• Masks: single-channel PNGs with per-pixel class IDs.
• Class map: id2label.json with IDs 0…47, where:
  • 0 = unlabelled (pixels without annotation)
  • 1…46 = DeepFashion2-style apparel & accessories
  • 47 = bikini (from Roboflow source)

There is no explicit background class. Unannotated pixels are 0 = unlabelled.

---

Splits

• train
• validation
• (optional) test

A multilabel-stratified split is used to keep class presence ratios similar across splits.

---

How to Load (Hugging Face Datasets)

from datasets import load_dataset

# Replace if your repo id differs
ds = load_dataset("Itbanque/fashion_segformer")

train_ds = ds["train"]
val_ds   = ds["validation"]
print(train_ds[0].keys())  # {'image', 'label'}

Each sample:

• image: PIL RGB image (H, W, 3)
• label: PIL "L" image (H, W), values in {0..47}

---

Quick Peek

import numpy as np
import matplotlib.pyplot as plt

ex = train_ds[0]
img, mask = ex["image"], ex["label"]

print("unique mask ids:", np.unique(np.array(mask)))

plt.figure(figsize=(8,4))
plt.subplot(1,2,1); plt.imshow(img); plt.axis("off"); plt.title("image")
plt.subplot(1,2,2); plt.imshow(mask, cmap="tab20"); plt.axis("off"); plt.title("mask (IDs)")
plt.tight_layout(); plt.show()

---

Class Map (excerpt)

{
  "0": "unlabelled",
  "1": "shirt, blouse",
  "2": "top, t-shirt, sweatshirt",
  "...": "...",
  "46": "tassel",
  "47": "bikini"
}

---

File Structure (example)

.
├─ README.md
├─ id2label.json
├─ fashion_segmentation_split_stats.png   # optional visual
└─ data files (parquet/arrow or image/mask folders)

---

Acknowledgements & Licenses

This dataset combines and repackages annotations and/or images from:

• Hugging Face dataset: sayeed99/fashion_segmentation
• Roboflow project: bikini-detector-v24

Please review the original sources’ licenses/terms of use. If you publish results, cite the sources appropriately.

---

Citation (example)

@misc{fashion-segmentation-combined,
  title  = {Fashion Segmentation (Combined DeepFashion2 + Bikini)},
  author = {Jian},
  year   = {2025},
  howpublished = {\url{https://huggingface.co/datasets/fj11/fashion}}
}