The dataset can be used to train instruction-guided image editing models.
It is built on top of InstructPix2Pix CLIP-filtered with new edit instructions.
For each sample, source_image is the original image, instruction is the edit instruction, target_image is the edited image, and original_instruction is the edit instruction from the InstructPix2Pix CLIP-filtered dataset.
Please refer to our repo to see how the edit instructions… See the full description on the dataset page:
https://huggingface.co/datasets/SherryXTChen/InstructCLIP-InstructPix2Pix-Data.
