Update README.md

README.md

@@ -38,74 +38,103 @@ pretty_name: Outfit Combos (Winter Wearability)
 size_categories:
 - n<1K
 ---
-This dataset contains 30+ original images of **shirt–pant outfit combinations**, each labeled by **formality** (casual, semi-formal, formal) and **winter wearability** (binary).  
-All images were manually collected and resized to 224×224.  
-An augmented split (300+ images) was generated using Torchvision transformations to support educational computer vision tasks.
+# Dataset Card for Outfit Dataset
 
-# **Model Card: Outfit Combos Dataset**
+This dataset card documents the **Clothing Outfit Dataset**.  
+It contains 30 original outfit images (shirt-pant combinations) with categorical and binary labels, with an augmented split expanding to 300 images.  
 
-## Purpose
-This dataset was created to explore how clothing combinations (shirt + pant) can be categorized by **formality** and whether they can be **geared for winter wear or not**.  
-The goal is educational: to practice dataset construction, augmentation, and deployment for computer vision tasks.
+## Dataset Details
 
+**Dataset Description**  
+- **Curated by:** Bareethul Kader (Carnegie Mellon University)  
+- **Language(s):** English (labels: formality, binary target)  
+- **License:** CC BY 4.0  
+- **Repository:** [bareethul/clothing-outfits](https://huggingface.co/datasets/bareethul/clothing-outfits)  
 
-## Composition
-- **30 original outfit images**, photographed by the creator.  
-- Each image resized to **224×224 px**.  
-- Labels include:  
-  - `Formality` (categorical: casual, semi formal, formal)  
-  - `is_winter_wearable` (binary: 1 = yes, 0 = no)  
+---
+
+## Uses
+
+**Direct Use**  
+- Educational practice in **computer vision dataset creation and augmentation**.  
+- Binary classification task: predict whether an outfit is **wearable in winter** with a jacket.  
+- Multi-class classification: predict outfit **formality** (Casual, Semi-Formal, Formal).  
+
+**Out-of-Scope Use**  
+- Not intended for fashion recommendations or e-commerce applications.  
+- Images represent a limited personal wardrobe, not generalizable fashion data.  
+
+---
+
+## Dataset Structure
+
+- **Original split:** 30 manually captured outfit images (shirt–pant combos).  
+- **Augmented split:** 300 images generated with **torchvision transforms** (random crop, flip, rotation, color jitter, etc.).  
+
+**Features:**  
+- `Image` (224×224 JPG, RGB)  
+- `Formality` (categorical: Casual, Semi-Formal, Formal)  
+- `is_winter_wearable` (binary target: `1 = yes, 0 = no`)  
+
+---
 
+## Dataset Creation
 
-## Collection
-- All images were taken manually from mine and my roommate’s wardrobes(not web sourced).  
-- Faces/identifiable people excluded; dataset only includes clothing objects.  
-- No sensitive or personal information included.  
+**Curation Rationale**  
+To study how outfit images can be labeled for **seasonal wearability** and **formality classification** in a simplified educational dataset.  
 
-## Preprocessing & Augmentation
-- **Preprocessing**: resized all images to 224×224.  
-- **Augmentation**: Torchvision transforms including:  
-  - Random crops and resizes  
-  - Horizontal/vertical flips  
-  - Random rotations (±15°)  
-  - Color jitter (brightness, contrast, saturation, hue)  
-  - Optional RandAugment for stronger diversity  
-- Expanded dataset from 30 images to **300 augmented images**.  
-- Augmentations were **label preserving**.  
+**Data Collection and Processing**  
+- Original images manually photographed by the dataset creator (shirt–pant combos).  
+- Images resized to **224×224 pixels** for consistency.  
+- Binary label `is_winter_wearable` assigned based on whether the outfit could be worn in winter with a jacket.  
+- Augmentation applied using PyTorch/Torchvision (random flips, rotations, color jitter, erasing, etc.).  
 
+**Source Data Producers**  
+- Original producer: Bareethul Kader (personal clothing images).  
+- Dataset curated solely for educational purposes.  
 
-## Labels
-- `Image`: RGB photo of outfit combo  
-- `Formality`: categorical label (casual/semi formal/formal)  
-- `is_winter_wearable`: binary label (1 = yes, 0 = no)  
+---
+
+## Annotations
+
+- **Annotation Process:**  
+  - `Formality` labeled manually by the creator (Casual, Semi-Formal, Formal).  
+  - `is_winter_wearable` labeled based on personal judgment.  
+- **Annotators:** Dataset creator.  
 
-## Splits
-- **original**: 30 manually captured outfit images  
-- **augmented**: 300 augmented images  
+---
+
+## Personal and Sensitive Information
 
+- No faces, people, or identifying information included.  
+- Only images of **inanimate clothing items**.  
 
-## Intended Use / Limits
-- **For use in**:  
-  - Educational purposes (image classification practice)  
-  - Demonstrating augmentation pipelines  
-- **Not for use in**:  
-  - Real fashion recommendation systems  
-  - Commercial retail applications  
+---
 
+## Bias, Risks, and Limitations
 
-## Ethical Notes
-- Dataset avoids sensitive content and privacy issues.  
-- Focuses only on clothing items; no personal data is included.  
+- Small dataset (30 images) — not representative of all outfit styles.  
+- Subjective labels (`formality`, `is_winter_wearable`) may reflect individual judgment, not universal agreement.  
+- Augmentation may create visually unrealistic variants.  
 
+---
+
+## Recommendations
+
+Users should be aware of the **limited scope and subjectivity** of this dataset.  
+It is intended only for **educational demonstrations** of image dataset creation and augmentation.  
+
+---
 
-## License
-- **CC BY 4.0** - free to share and adapt with attribution.  
+## Citation
 
+**BibTeX:**  
+```bibtex
+@dataset{bareethul_clothing_outfits,
+  author       = {Kader, Bareethul},
+  title        = {Clothing Outfit Dataset},
+  year         = {2025},
+  publisher    = {Hugging Face Datasets},
+  url          = {https://huggingface.co/datasets/bareethul/clothing-outfits}
+}
 
-## AI Usage Disclosure
-- ChatGPT was used as a coding teammate for:  
-  - Defining the augmentation approach  
-  - Helping create image augmentation code blocks  
-  - Building the contact sheet visualization  
-  - Polishing final notebooks and this model card  
-  - All images were manually collected by the creator.