OpenHay/zalo-ai-2025-training-data-v2

Zalo AI Challenge 2025 - RoadBuddy Training Data V2set

This dataset contains training data v2 for the RoadBuddy – Understanding the Road through Dashcam AI challenge from Zalo AI Challenge 2025.

Dataset Description

The challenge aims to build a driving assistant capable of understanding video content from dashcams to quickly answer questions about traffic signs, signals, and driving instructions in Vietnam.

Dataset Structure

Files

• frames/: Directory containing 3,843 extracted frame images from dashcam videos

  • Format: JPG images
  • Naming convention: {video_id}_{clip_info}_frame_{frame_number}.jpg
  • 7 frames per video clip

• qwen3vl_train_v2.json: Training Data V2 in conversation format

  • Format: JSON array of samples
  • Each sample contains:
    • id: Unique identifier
    • image: List of 7 frame paths
    • conversations: Question-answer pairs in Vietnamese with detailed analysis

Data Statistics

• Total samples: ~1490+ QA pairs
• Total frames: 3,843 images
• Language: Vietnamese
• Video duration: 5-15 seconds per clip
• Scenarios: Urban/highway traffic, day/night, various weather conditions

Sample Structure

{
  "id": "sample_0001",
  "image": [
    "frames/video_id_frame_0.jpg",
    "frames/video_id_frame_1.jpg",
    ...
  ],
  "conversations": [
    {
      "from": "human",
      "value": "<image>\n<image>...\nQUESTION:\n...\nANSWER CHOICES:\n..."
    },
    {
      "from": "gpt",
      "value": "Phân tích...\nANSWER: A"
    }
  ]
}

Usage

Loading the Dataset

from datasets import load_dataset
from huggingface_hub import hf_hub_download
import json

# Load the JSON file
json_path = hf_hub_download(
    repo_id="OpenHay/zalo-ai-2025-training-data-v2",
    filename="qwen3vl_train_v2.json",
    repo_type="dataset"
)

with open(json_path, 'r', encoding='utf-8') as f:
    data = json.load(f)

# Access frames
for sample in data:
    sample_id = sample['id']
    frames = sample['image']
    for frame_path in frames:
        frame_file = hf_hub_download(
            repo_id="OpenHay/zalo-ai-2025-training-data-v2",
            filename=frame_path,
            repo_type="dataset"
        )
        # Process frame...

Dataset Details

Question Types

The dataset includes questions about:

• Traffic signs recognition
• Traffic signal states
• Lane markings and arrows
• Road directions and navigation
• Speed limits and restrictions
• Traffic rules compliance
• Vietnamese traffic regulations

Frame Extraction

Frames are extracted at regular intervals from dashcam video clips to capture the temporal progression of traffic scenarios.

Citation

If you use this dataset, please cite:

@dataset{zalo_ai_2025_roadbuddy,
  title={RoadBuddy Training Data V2set - Zalo AI Challenge 2025},
  author={OpenHay},
  year={2025},
  publisher={HuggingFace},
  howpublished={\url{https://huggingface.co/datasets/OpenHay/zalo-ai-2025-training-data-v2}}
}

License

This dataset is released under CC-BY-NC-4.0 license for non-commercial use only, as per the Zalo AI Challenge 2025 competition rules.

Competition Information

• Challenge: RoadBuddy – Understanding the Road through Dashcam AI
• Organizer: Zalo AI
• Year: 2025
• Task: Visual Question Answering on Vietnamese traffic dashcam videos
• Constraints:
  • Model size ≤ 8B parameters
  • Inference time ≤ 30s/sample
  • Vietnamese traffic law compliance required

Contact

For questions about this dataset, please contact the repository maintainers.