DRL4Real Dataset
Welcome to the DRL4Real dataset download page. Our dataset is specifically designed for disentangled representation learning and controllable generation, covering multiple realistic scenarios and diverse disentanglement factors. It consists of three main categories as shown below.
Single-Factor Dataset
This dataset contains over 13,000 images with single-factor variations across 13 different categories including car, cat, cattle, dog, eagle, facial, flower, fruit, horse, pigeon, tree, vegetable, and time-lapse photography.
Dataset Details:
- Size: 13,000+ images
- Categories: 13 distinct categories
- Format: High-resolution images with detailed text descriptions
- Features: Controlled single-factor variations with corresponding text descriptions
Multi-Factor Dataset
This dataset focuses on multi-factor variations across the same 12 categories as the single-factor dataset. It provides only text descriptions showing how multiple factors can be manipulated simultaneously.
Dataset Details:
- Size: 24000+ texts
- Categories: 12 distinct categories
- Format: Detailed text descriptions
- Features: Multiple attribute variations with controlled factor changes
Autonomous Driving Dataset
A vehicle-centered dataset featuring 5 camera perspectives with simultaneous variations in 8 key factors including vehicle speed, weather conditions, lighting, and other environmental factors.
Dataset Details:
- Size: 20,000 high-quality images
- Camera Views: 5 perspectives (4 outward-facing + 1 top-down)
- Vehicles: 3 distinct vehicle types
- Scenarios: 5 background scenarios
- Factors: 8 key variation factors
- Format: 100 frames per sample