PIE-Bench
Prompt-based Image Editing Benchmark
ImagesTextshttps://docs.google.com/document/d/1K7pjAgGwS8-lsA-D6Me4jAKmYszKpcyydiAEyZuTm2I/edit?usp=sharingIntroduced 2023-10-02
PIE-Bench comprises 700 images featuring 10 distinct editing types. Images are evenly distributed in natural and artificial scenes (e.g., paintings) among four categories: animal, human, indoor, and outdoor. Each image in PIE-Bench includes five annotations: source image prompt, target image prompt, editing instruction, main editing body, and the editing mask. Notably, the editing mask annotation (indicating the anticipated editing region) is crucial in accurate metrics computations as we expect the editing to only occur within a designated area.
Benchmarks
1 Image, 2*2 Stitchi/Background PSNR1 Image, 2*2 Stitchi/Background LPIPS1 Image, 2*2 Stitchi/Structure Distance1 Image, 2*2 Stitchi/CLIPSIM10-shot image generation/Background PSNR10-shot image generation/Background LPIPS10-shot image generation/Structure Distance10-shot image generation/CLIPSIMImage Generation/Background PSNRImage Generation/Background LPIPSImage Generation/Structure DistanceImage Generation/CLIPSIMText-to-Image Generation/Background PSNRText-to-Image Generation/Background LPIPSText-to-Image Generation/Structure DistanceText-to-Image Generation/CLIPSIM