Age Estimation on UTKFace
4.15MAEDLDL-v2 (ThinAgeNet)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DLDL-v2 (ThinAgeNet)External Data=false, Single Model=true, Network=Thin-VGG, Scope=Entire (0-116 years)2020.07 | 4.15 | — | |
| DLDL-v2 (TinyAgeNet)External Data=false, Single Model=true, Network=Tiny-VGG, Scope=Entire (0-116 years)2020.07 | 4.26 | — | |
| Supervised BaselineType=Fully Supervised2026.03 | 4.42 | — | |
| Axel Berg et al.External Data=false, Single Model=true, Network=ResNet-50, Venue=ICPR 20192020.07 | 4.55 | — | |
| Gustafsson et al.External Data=false, Single Model=true, Network=ResNet-50, Venue=ECCV 20202020.07 | 4.65 | — | |
| MobileAgeNetSetting=Full range, stratified split, Params=3.23M, GFLOPs=0.233G2026.04 | 4.65 | — | |
| Andrey SavchenkoExternal Data=true, Single Model=true, Network=MobileNet-v2, Venue=PeerJ 20192020.07 | 5.44 | — | |
| Savchenko MobileNetSetting=Full range, fine-tuned, Params=not reported, GFLOPs=not reported2026.04 | 5.44 | — | |
| CORALExternal Data=false, Single Model=true, Network=ResNet-50, Venue=PRL 20202020.07 | 5.47 | — | |
| MobileNet-v1Param count, MB=3.22021.03 | 5.74 | — | |
| OR-CNNExternal Data=false, Single Model=true, Network=ResNet-50, Venue=CVPR 20162020.07 | 5.74 | — | |
| Savchenko MobileNetSetting=Full range, VGGFace2 pretrain, Params=not reported, GFLOPs=not reported2026.04 | 5.74 | — | |
| LSDK=12026.03 | 5.9 | — | |
| kNNK=12026.03 | 5.98 | — | |
| Gemma 3 4B-itStrategy=0-Shot2026.03 | 6.1 | — | |
| RandomK=12026.03 | 6.14 | — | |
| LSDK=42026.03 | 6.27 | — | |
| VGG-16 (DEX)Param count, MB=262.52021.03 | 6.48 | — | |
| LSDK=162026.03 | 6.64 | — | |
| LSDK=82026.03 | 7.05 | — | |
| kNNK=42026.03 | 7.27 | — | |
| MobileNet v2 (Agegendernet)Param count, MB=7.12021.03 | 7.29 | — | |
| kNNK=162026.03 | 7.6 | — | |
| kNNK=82026.03 | 7.61 | — | |
| New model from Insight FaceParam count, MB=0.32021.03 | 8.44 | — | |
| ResNet-50 (Insight Face)Param count, MB=60.12021.03 | 8.57 | — | |
| FaceNetParam count, MB=12.32021.03 | 8.58 | — | |
| RandomK=42026.03 | 10.66 | — | |
| RandomK=162026.03 | 12.51 | — | |
| RandomK=82026.03 | 14.86 | — | |
| MiVOLO2026.01 | — | 4.23 | |
| Qwen2.5-VL-3BBackbone=Qwen2.5-VL-3B2026.01 | — | 6.95 | |
| Qwen2.5-VL-3B + RARLBackbone=Qwen2.5-VL-3B, Training=RARL2026.01 | — | 4.02 | |
| Qwen2.5-VL-3B + SFTBackbone=Qwen2.5-VL-3B, Training=SFT2026.01 | — | 5.51 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.01 | — | 6.14 | |
| Qwen2.5-VL-7B + RARLBackbone=Qwen2.5-VL-7B, Training=RARL2026.01 | — | 3.81 | |
| Qwen2.5-VL-7B + SFTBackbone=Qwen2.5-VL-7B, Training=SFT2026.01 | — | 5.02 | |
| RankingCLIP2026.01 | — | 3.83 |