Image Captioning on Sydney Captions
79.3BLEU-4Ours
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OursVisual Encoder=CLIP-Cap-4, Text Decoder=LlamaV22025.12 | 79.3 | — | 78.9 | 48.2 | 2.864 | 92.7 | 87.2 | 82.6 | 45.8 | |
| SkyEyeGPTVisual Encoder=EVA-G, Text Decoder=LlamaV2-Chat2025.12 | 77.4 | — | 77.7 | 46.6 | 1.811 | 91.9 | 85.6 | 80.9 | — | |
| SkyEyeGPTTotal Params=7B2025.12 | 77.4 | — | — | — | 1.811 | 91.9 | — | — | — | |
| Def. Trans.Total Params=90M*2025.12 | 66.6 | — | — | — | 3.037 | 83.7 | — | — | — | |
| Aware-Trans.Total Params=28M*2025.12 | 65.1 | — | — | — | 2.832 | 85.4 | — | — | — | |
| EarthDialCategory=Open-source Remote Sensing Vision-Language Models2025.12 | 64.04 | — | 43.75 | 54.91 | 120.9 | — | — | — | — | |
| RSGPTVisual Encoder=EVA-G, Text Decoder=Vicuna2025.12 | 62.2 | — | 74.8 | 41.4 | 2.731 | 82.3 | 75.3 | 68.6 | — | |
| RSGPTTotal Params=13B2025.12 | 62.2 | — | — | — | 2.731 | 82.3 | — | — | — | |
| HCNetTotal Params=25M*2025.12 | 61 | — | — | — | 2.471 | 76.9 | — | — | — | |
| MC-NetTotal Params=138M*2025.12 | 60.7 | — | — | — | 2.564 | 83.4 | — | — | — | |
| MLCA-NETVisual Encoder=VGG16, Text Decoder=LSTM2025.12 | 58 | — | 71.1 | 39 | 2.324 | 83.1 | 74.2 | 65.9 | 40.9 | |
| MLCA-NETTotal Params=138M*2025.12 | 58 | — | — | — | 2.324 | 83.1 | — | — | — | |
| GeoMELTTotal Params=271M2025.12 | 57.5 | — | — | — | 2.643 | 79.1 | — | — | — | |
| RS-CapRetVisual Encoder=CLIP-Cap-4, Text Decoder=LlamaV2, Finetuned=true2025.12 | 56.4 | — | 70.7 | 38.8 | 2.392 | 78.7 | 70 | 62.8 | 43.4 | |
| MF-RSVLMCategory=Open-source Remote Sensing Vision-Language Models2025.12 | 56.21 | — | 71.85 | 72.86 | 242.48 | — | — | — | — | |
| RS-CapRetVisual Encoder=CLIP-Cap-4, Text Decoder=LlamaV2, Finetuned=false2025.12 | 54.5 | — | 70.4 | 38.3 | 2.39 | 78.2 | 68.8 | 61.1 | 42.3 | |
| RS-CapRetTotal Params=7B2025.12 | 54.5 | — | — | — | 2.39 | 78.2 | — | — | — | |
| VHMCategory=Open-source Remote Sensing Vision-Language Models2025.12 | 44.67 | — | 23.76 | 35.11 | 67.5 | — | — | — | — | |
| SkySenseGPTCategory=Open-source Remote Sensing Vision-Language Models2025.12 | 42.47 | — | 21.51 | 24.95 | 52.58 | — | — | — | — | |
| Gemini-2.0Category=Closed-source Commercial Vision-Language Models2025.12 | 31.41 | — | 16.99 | 24.17 | 38.76 | — | — | — | — | |
| Qwen2.5-VLCategory=Open-source Vision-Language Models2025.12 | 28.6 | — | 16.87 | 18.77 | 31.81 | — | — | — | — | |
| GPT-5Category=Closed-source Commercial Vision-Language Models2025.12 | 28.5 | — | 17.47 | 24.48 | 39.09 | — | — | — | — | |
| MiniGPT-v2Category=Open-source Vision-Language Models2025.12 | 26.17 | — | 12.3 | 17.03 | 23.55 | — | — | — | — | |
| Claude-sonnet-4Category=Closed-source Commercial Vision-Language Models2025.12 | 19.85 | — | 12.52 | 20.14 | 27.55 | — | — | — | — |