Image Captioning on COCO (B@4, CIDEr)
152.7CIDErGIT2 (5.1B)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GIT2 (5.1B)#Example=12.9B, Reinforcement learning=true2023.04 | 152.7 | — | 26.4 | |
| VALORL#Example=33.5M, Resolution=392, Reinforcement learning=true2023.04 | 152.5 | — | 25.7 | |
| GIT#Example=800M, Reinforcement learning=true2023.04 | 151.1 | — | 26.3 | |
| PALI (16.9B)#Example=1.6B, Reinforcement learning=false2023.04 | 149.1 | — | — | |
| LEMON#Example=200M, Reinforcement learning=true2023.04 | 145.5 | — | 25.5 | |
| GITL#Example=20M, Reinforcement learning=true2023.04 | 144.6 | — | 25.4 | |
| CoCa#Example=4.8B, Reinforcement learning=false2023.04 | 143.6 | — | 24.7 | |
| SimVLM#Example=1.8B, Reinforcement learning=false2023.04 | 143.3 | — | 25.4 | |
| Oscar#Example=10M, Reinforcement learning=true2023.04 | 140.9 | — | 25.2 | |
| VinVL#Example=10M, Reinforcement learning=true2023.04 | 140 | — | 24.5 | |
| Flamingo (80B)#Example=2.3B, Reinforcement learning=false2023.04 | 138.1 | — | — | |
| SQ-LLAVAEvaluation Protocol=Fine-tune2024.03 | 136.78 | 40.76 | — | |
| BLIP#Example=135M, Reinforcement learning=false2023.04 | 136.7 | — | — | |
| UFO#Example=10M, Reinforcement learning=false2023.04 | 131.2 | — | 23.3 | |
| MAPLEvaluation Protocol=Fine-tune2024.03 | 125.2 | 36.45 | — | |
| GiT-H universal#Params=756M2024.03 | 124.2 | 37.7 | — | |
| Uni-Perceiver-v2 LARGEparameters=446M, task-specific fine-tuning=false2022.11 | 122.5 | 36.5 | — | |
| Unified-IO XLparameters=2.9B, task-specific fine-tuning=false2022.11 | 122.3 | — | — | |
| GiT-L universal#Params=387M2024.03 | 117.5 | 36.2 | — | |
| Uni-Perceiver-v2 BASEparameters=308M, task-specific fine-tuning=false2022.11 | 116.9 | 35.4 | — | |
| UniTabparameters=185M, task-specific fine-tuning=false2022.11 | 115.8 | — | — | |
| eP-ALMEvaluation Protocol=Fine-tune2024.03 | 111.63 | 33.35 | — | |
| LLaVA-v1.5Evaluation Protocol=Zero-shot2024.03 | 111.46 | 29.96 | — | |
| SQ-LLAVAEvaluation Protocol=Zero-shot2024.03 | 110.77 | 29.85 | — | |
| ClipCapEvaluation Protocol=Fine-tune2024.03 | 108.55 | 32.6 | — | |
| GiT-B universal#Params=131M2024.03 | 108.3 | 33.6 | — | |
| DiscriTuneEvaluation Protocol=Fine-tune2024.03 | 105.4 | 32.31 | — | |
| eP-ALMEvaluation Protocol=Zero-shot2024.03 | 97.22 | 29.47 | — | |
| MAPLEvaluation Protocol=Zero-shot2024.03 | 54.3 | 12.3 | — | |
| DiscriTuneEvaluation Protocol=Zero-shot2024.03 | 53.2 | 13.99 | — | |
| ClipCapEvaluation Protocol=Zero-shot2024.03 | 37.03 | 8.5 | — | |
| Pix2Seq v2parameters=132M, task-specific fine-tuning=false2022.11 | — | 34.9 | — | |
| Uni-Perceiver BASEparameters=124M, task-specific fine-tuning=false2022.11 | — | 32 | — | |
| Uni-Perceiver LARGEparameters=354M, task-specific fine-tuning=false2022.11 | — | 35.3 | — | |
| Uni-Perceiver-MOE BASEparameters=167M, task-specific fine-tuning=false2022.11 | — | 33.2 | — | |
| Uni-Perceiver-MOE LARGEparameters=505M, task-specific fine-tuning=false2022.11 | — | 35.5 | — |