This dataset is designed for fine-tuning the CLIP model on Spectrum icons, enabling enhanced visual and semantic understanding of Adobe Spectrum assets.
Purpose: Fine-tuning the CLIP model on Spectrum icons for improved icon retrieval and classification based on visual and abstract meanings.
Data Structure: Each row includes:
file_path: The relative path to the image.
caption: A detailed description of… See the full description on the dataset page:
https://huggingface.co/datasets/JianLiao/spectrum-icons.
