---
dataset_info:
  features:
    - name: image
      dtype: image
    - name: prompt
      dtype: string
    - name: ground_truth
      dtype: string
license: other
task_categories:
  - image-classification
  - visual-question-answering
tags:
  - fashion
  - clothing
  - garment
  - apparel
  - vlm
  - multimodal
size_categories:
  - 10K<n<100K
---

# Denali-AI Train Set 35k (Strict V2)

A curated training dataset of **34,493 garment images** with structured attribute annotations for fine-tuning Vision-Language Models on apparel classification.

## Dataset Description

Each sample contains:
- **image**: A garment photograph (clothing laid flat, on hangers, or worn)
- **prompt**: Classification instruction prompt for VLM inference
- **ground_truth**: JSON string with 9 garment attributes

### Attribute Schema

| Field | Description | Examples |
|-------|-------------|----------|
| type | Garment type | Shirt, Pants, Dress, Jacket |
| color | Primary color | Blue, Red, Black |
| pattern | Fabric pattern | Solid, Striped, Floral, Plaid |
| neckline | Neckline style | Crew, V-Neck, Collared Neckline |
| sleeve_length | Sleeve length | Short, Long, Sleeveless |
| closure | Closure type | Button, Zipper, Pullover |
| brand | Brand (if visible on tag) | Nike, Gap, N/A |
| size | Size (if visible on tag) | M, XL, N/A |
| defect_type | Visible defects | Hole, Stain, N/A |

## Data Sources

Images sourced from multiple collections:
- **crowd**: Crowdsourced garment photos
- **posh**: Resale marketplace listings
- **zen**: Processing facility station captures
- **nike-shoes-authenticity**: Nike footwear samples
- **training-eval**: Held-out evaluation subset

## Usage



## Intended Use

This dataset is intended for SFT/GRPO training of VLM models on garment attribute extraction. It is disjoint from the [Denali-AI/eval-benchmark-18k](https://huggingface.co/datasets/Denali-AI/eval-benchmark-18k) evaluation set.

## Organization

Part of the [Denali-AI](https://huggingface.co/Denali-AI) garment classification project.