Visual Instruction Generation on WikiHow-VGSI (test)
84Step FaithfulnessStable Diffusion
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Stable Diffusionsetup=zero-shot, conditioned_on_input_image=false2024.12 | 84 | 3 | |
| ShowHowTosetup=zero-shot2024.12 | 72 | 12 | |
| GenHowTosetup=zero-shot, generation_strategy=iterative2024.12 | 60 | 6 | |
| Source sequences2024.12 | 60 | 100 | |
| StackedDiffusionsetup=zero-shot, generation_strategy=input masking2024.12 | 57 | 7 | |
| Phung et al.setup=zero-shot, backbone=Stable Diffusion [48], generation_strategy=input masking2024.12 | 46 | 4 | |
| AURORAsetup=zero-shot, generation_strategy=iterative2024.12 | 33 | 15 | |
| InstructPix2Pixsetup=zero-shot, generation_strategy=iterative2024.12 | 32 | 12 | |
| Random2024.12 | 26 | 0 | |
| Copy of the input image2024.12 | 26 | 26 |