PrismLayersPro100K — Multi-layer image dataset for transparent image generation

High-quality image dataset containing 100,000 multi-layered samples, including foreground objects and background scenes, with embedded text for the controlled generation of transparent images.

Download dataset

Size

100,000 multi-layered images, PNG formats with transparent layers

Licence

Apache 2.0

Description

‍

PrismLayersPro 100K is a transparent multi-layered image dataset designed to train and evaluate generative models that can compose images with multiple layers, including objects, text, and backgrounds. Each image is structured in layers allowing precise control over the composition.

‍

What is this dataset for?

‍

Train multi-layered generative models for the composition of transparent images
Study the decomposition and recomposition of complex images with superimposed layers
Develop applications for generating images that integrate text in a contextual manner

‍

Can it be enriched or improved?

‍

Yes, you can enrich this dataset by adding annotations on objects or types of text, by diversifying the scenes or by integrating additional formats for better flexibility.

‍

🔎 In summary

Criterion	Evaluation
🧩 Ease of use	⭐⭐⭐⭐✩ (Standard multi-layer PNG format, easy to handle with appropriate tools)
🧼 Need for cleaning	⭐⭐⭐⭐⭐ (Low, ready-to-use dataset)
🏷️ Annotation richness	⭐⭐⭐✩✩ (Implicit annotations via multiple layers, to be completed if needed)
📜 Commercial license	✅ Yes (Apache 2.0)
👨‍💻 Beginner friendly	⚠️ Moderate, requires knowledge in multi-layer image processing
🔁 Fine-tuning ready	🎯 Perfect for GANs and advanced generative models
🌍 Cultural diversity	🌐 Visual dataset, high visual diversity