GLIGEN: Open-Set Grounded Text-to-Image Generation

Benchmark Model Rank Results
layout-to-image-generation-on-layoutbench-coco-combinationGLIGEN#2AP: 36.3
layout-to-image-generation-on-layoutbench-coco-numberGLIGEN#2AP: 30.7
layout-to-image-generation-on-layoutbench-coco-positionGLIGEN#2AP: 38.9
layout-to-image-generation-on-layoutbench-coco-sizeGLIGEN#2AP: 33.3
conditional-text-to-image-synthesis-on-coco-migGligen (zero-shot)#3instance success rate: 0.30mIoU: 0.27
text-to-image-generation-on-cocoGLIGEN (fine-tuned, Detection + Caption data)#3FID: 5.61
text-to-image-generation-on-cocoGLIGEN (fine-tuned, Detection data only)#4FID: 5.82
text-to-image-generation-on-cocoGLIGEN (fine-tuned, Grounding data)#7FID: 6.38