Handwriting Retrieval on Handwriting Spanish synthetic disjoint fonts (Out-of-Domain (OOD))
86.05Top-1 AccuracyOurs
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| OursFine-tuning=true, Paras(M)=1.29, Latency(ms)=2.892026.01 | 86.05 | 95.37 | 96.93 | 90.94 | 89.03 | |
| InternVL3.5-8BParas(M)=8528.32, Latency(ms)=188.522026.01 | 84.84 | 88.48 | 89.57 | 87.06 | 62.61 | |
| QWen3-VL-4BParas(M)=4437.82, Latency(ms)=18.212026.01 | 84.44 | 87.12 | 87.76 | 85.97 | 85.02 | |
| MiniCPM-o 2.6Paras(M)=8099.42, Latency(ms)=125.132026.01 | 82.3 | 85.86 | 86.91 | 84.36 | 82.84 | |
| GME-Qwen2VL-7BParas(M)=7746.38, Latency(ms)=197.232026.01 | 78.02 | 87.29 | 89.4 | 83.4 | 83.96 | |
| GME-Qwen2VL-2BParas(M)=2208.99, Latency(ms)=189.432026.01 | 72.26 | 84.16 | 87.05 | 79.07 | 77.97 | |
| EasyOCRParas(M)=30.10, Latency(ms)=20.332026.01 | 60.44 | 66.85 | 68.48 | 64.48 | 59.82 | |
| ABINetParas(M)=36.86, Latency(ms)=14.212026.01 | 58.79 | 61.34 | 61.8 | 60.23 | 58.67 | |
| RapidOCRParas(M)=15.00, Latency(ms)=96.102026.01 | 58.63 | 62.58 | 63.44 | 61.25 | 60.49 | |
| PARSeqParas(M)=6.02, Latency(ms)=10.752026.01 | 57.61 | 60.88 | 61.65 | 59.45 | 57.91 | |
| SigLIP 2 GiantParas(M)=1871.39, Latency(ms)=23.792026.01 | 55.26 | 62.86 | 65.48 | 60.46 | 61.48 | |
| Janus-Pro-7BParas(M)=7420.43, Latency(ms)=572.832026.01 | 51.53 | 55.16 | 56.62 | 53.87 | 53.9 | |
| TrOCRParas(M)=333.92, Latency(ms)=29.742026.01 | 51.52 | 55.82 | 57.03 | 54.05 | 51.66 | |
| SuryaParas(M)=38.42, Latency(ms)=96.702026.01 | 49.9 | 52.99 | 54.01 | 51.77 | 50.05 | |
| SigLIP 2 So400mParas(M)=1135.67, Latency(ms)=27.232026.01 | 48.32 | 54.89 | 56.83 | 52.86 | 54.63 | |
| Llama-3.2-11B-VisionParas(M)=10670.22, Latency(ms)=717.802026.01 | 48.28 | 49.97 | 50.35 | 49.23 | 49.7 | |
| Phi-3.5-VisionParas(M)=4146.62, Latency(ms)=854.852026.01 | 47.4 | 50.91 | 52 | 49.46 | 50.52 | |
| Chinese CLIP ViT LargeParas(M)=406.23, Latency(ms)=20.602026.01 | 45.97 | 55.7 | 59.77 | 53.02 | 55.15 | |
| Chinese CLIP ViT HugeParas(M)=957.60, Latency(ms)=20.632026.01 | 45.8 | 54.51 | 58.25 | 52.27 | 54.07 | |
| SigLIP 2 LargeParas(M)=881.53, Latency(ms)=26.572026.01 | 45.07 | 53.07 | 55.63 | 50.4 | 52.79 | |
| Ours w/o FineTuningFine-tuning=false, Paras(M)=1.29, Latency(ms)=2.892026.01 | 43.03 | 52.02 | 55.58 | 49.53 | 49.94 | |
| Florence-2-Large (OCR)Paras(M)=776.47, Latency(ms)=159.092026.01 | 41.13 | 44.34 | 45.09 | 43.57 | 45.02 | |
| Nomic-Embed-VisionParas(M)=229.68, Latency(ms)=4.032026.01 | 38.82 | 45.8 | 47.66 | 43.87 | 46.04 | |
| Chinese CLIP ViT BaseParas(M)=188.26, Latency(ms)=26.352026.01 | 36.56 | 45.59 | 49.78 | 43.41 | 45.84 | |
| AltCLIPParas(M)=864.19, Latency(ms)=26.852026.01 | 32.45 | 39.47 | 42.14 | 37.75 | 40.02 | |
| SigLIP 2 ViT-BParas(M)=375.23, Latency(ms)=27.672026.01 | 31.13 | 42.18 | 45.52 | 38.4 | 41.12 | |
| MM-EMBEDParas(M)=8175.51, Latency(ms)=145.172026.01 | 29.77 | 38.69 | 41.88 | 36.15 | 38.07 | |
| DeepSeek-OCRParas(M)=3336.11, Latency(ms)=719.442026.01 | 26.96 | 33.39 | 36.1 | 31.96 | 31.64 | |
| E5-VParas(M)=8355.28, Latency(ms)=232.672026.01 | 18.53 | 24.42 | 29.37 | 24.79 | 26.45 | |
| BGE-VisualizedParas(M)=872.91, Latency(ms)=34.132026.01 | 10.39 | 16.98 | 20.91 | 16.12 | 19.21 |