Text-to-Image Generation on MSCOCO 30K 2014 (val)
10.65FIDDALL-E2
Evaluation Results
| Method | Links | |
|---|---|---|
| DALL-E2Datasize (M)=2502025.02 | 10.65 | |
| CS-Aligner + SD-unclip decoderDatasize (M)=0.08(COCO), Decoder=SD-unclip2025.02 | 10.88 | |
| CS-Aligner + Karlo decoderDatasize (M)=0.08(COCO), Decoder=Karlo2025.02 | 11.27 | |
| CS-Aligner + Kandinsky decoderDatasize (M)=0.08(COCO), Decoder=Kandinsky2025.02 | 12.62 | |
| SD-unclip v2.1Datasize (M)=20002025.02 | 13.15 | |
| SD v2.1Datasize (M)=20002025.02 | 14.51 | |
| Eclipse + Kandinsky decoderDatasize (M)=0.08(COCO), Decoder=Kandinsky2025.02 | 16.53 | |
| KandinskyDatasize (M)=1772025.02 | 20.48 | |
| KarloDatasize (M)=1152025.02 | 20.64 | |
| WurstchenDatasize (M)=14202025.02 | 23.6 | |
| Eclipse + Karlo decoderDatasize (M)=0.08(COCO), Decoder=Karlo2025.02 | 23.67 | |
| IB + Kandinsky decoderDatasize (M)=0.08(COCO), Decoder=Kandinsky2025.02 | 150.52 |