FusionAnything

# FusionAnthing_seedream ## Dataset Description This dataset contains **96,266** samples for image editing tasks, organized into 2 splits. ## Splits | Split | Samples | |-------|---------| | seedream_4imgs | 48,805 | | seedream_5imgs | 47,461 | ## Dataset Structure Each sample contains: - `input_images`: List of input image paths (relative to dataset root) - `output_image`: Output image path (relative to dataset root) - `instruction`: Text instruction/prompt for the editing task ## Usage ### Load All Splits ```python from datasets import load_dataset # Load the full dataset dataset = load_dataset("FusionAnthing_seedream") # Access specific splits train_data = dataset["unipic3_2imgs"] ``` ### Load Specific Split ```python from datasets import load_dataset # Load only one split dataset = load_dataset("FusionAnthing_seedream", split="unipic3_2imgs") ``` ### View Images ```python from datasets import load_dataset from PIL import Image dataset = load_dataset("FusionAnthing_seedream", split="unipic3_2imgs") # Get first sample sample = dataset[0] print(f"Instruction: {sample['instruction']}") print(f"Input images: {sample['input_images']}") print(f"Output image: {sample['output_image']}") ``` ## File Structure ``` FusionAnthing_seedream/ ├── README.md ├── data/ │ ├── unipic3_2imgs.parquet │ ├── unipic3_3imgs.parquet │ └── ... └── images/ ├── abc123.png └── ... ``` ## License Apache 2.0