HoagMin/Vietnamese_StreetFood_Dataset

Vietnamese Street Food Dataset

This dataset contains approximately 3,500 real-world images of popular Vietnamese street food dishes. The images were collected to support tasks such as image classification, food recognition, and cultural heritage preservation related to Vietnamese cuisine.

All photos focus on authentic street-style presentations — from markets, vendors, and everyday eating spots across Vietnam. The dataset is particularly useful for training models to identify Vietnamese foods in natural, varied conditions (different angles, lighting, backgrounds, and plating styles).

Key Features

• Total images: ~3,500
• Classes: 14 distinct Vietnamese street food dishes
• Indexing: Standard 0–13 labels
• Splits: train / validation / test (approximately 40% / 30% / 30%)
• Image quality: Real-world street photos — variable lighting, angles, and compositions for better generalization

Class Mapping (0-13 indexing)

ID	English Name	Vietnamese Name
0	Steamed Rice Rolls	Bánh cuốn
1	Vietnamese pyramid rice dumpling	Bánh giò
2	Grilled Pork Banh Mi	Bánh mì
3	Hue-style Spicy Beef Noodle Soup	Bún bò Huế
4	Grilled Pork with Vermicelli	Bún chả
5	Vermicelli with Tofu	Bún đậu mắm tôm
6	Grilled Pork Vermicelli	Bún thịt nướng
7	Pork Offal Porridge	Cháo lòng
8	Flattened green rice	Cốm
9	Broken Rice	Cơm tấm
10	Fried Rice	Cơm rang
11	Beef Pho	Phở bò
12	Spring Rolls	Gỏi cuốn
13	Sticky Rice	Xôi

Dataset Structure

After downloading and unzipping, the directory structure is organized for YOLO training:

dataset/
├── data.yaml       # Configuration file (Classes & Paths)
├── train/
│   ├── images/     # Training images
│   └── labels/     # Training labels (.txt)
├── valid/
│   ├── images/     # Validation images
│   └── labels/     # Validation labels (.txt)
└── test/
    ├── images/     # Test images
    └── labels/     # Test labels (.txt)