README.md

# Dummy Video Dataset

This is a dummy dataset created for testing video understanding models.

## Dataset Structure

Each sample contains:
- `video_name`: Name of the video file
- `duration`: Duration in seconds
- `question`: Multiple choice question
- `candidates`: List of possible answers
- `answer`: Correct answer (A, B, C, or D)
- `task_type`: Type of task (needle, counting, temporal, spatial, reasoning)
- `question_id`: Unique question identifier

## Usage

```python
from datasets import load_dataset

# Load dataset
dataset = load_dataset("ragavsachdeva/dummy-dataset")

# Access video files
# Video files are stored in the 'videos/' directory
```

## Video Files

The videos are simple generated content with:
- Colored backgrounds that change over time
- Moving white circles
- Frame counter text
- Duration ranging from 5-30 seconds

Total samples: 20