Monocular Depth Estimation on Cityscapes (test)
5.14RMSEFree Sinewich
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Free SinewichBackbone=Swin-Tiny [26], Pre-trained=ImageNet-22K [7], Decoder=HRNet2026.03 | 5.14 | — | — | — | — | — | — | 5.45 | 5.92 | |
| Flex-X-Large (Ours)Year=2026, T.F.=1, Data=M, FLOPs.(G)=19.62026.07 | 5.268 | 0.086 | 0.877 | 0.137 | 92.6 | 98.2 | 99.3 | — | — | |
| Flex-Large (Ours)Year=2026, T.F.=1, Data=M, FLOPs.(G)=9.22026.07 | 5.31 | 0.087 | 0.911 | 0.139 | 92.4 | 98.1 | 99.3 | — | — | |
| Flex-Medium (Ours)Year=2026, T.F.=1, Data=M, FLOPs.(G)=8.02026.07 | 5.358 | 0.089 | 0.885 | 0.143 | 91.7 | 97.9 | 99.3 | — | — | |
| TADFormerBackbone=Swin-Tiny [26], Pre-trained=ImageNet-22K [7], Decoder=HRNet2026.03 | 5.39 | — | — | — | — | — | — | 3.44 | 5.97 | |
| DSI-MonoViTYear=2025, T.F.=1, Data=M, FLOPs.(G)=47.72026.07 | 5.41 | 0.088 | 0.872 | 0.141 | 91.8 | 98.1 | 99.3 | — | — | |
| ProDepthTest frames=2 (-1, 0), Semantics=false, W x H=416 x 1282024.07 | 5.531 | 0.095 | 0.876 | 0.146 | 90.8 | 97.8 | 99.3 | — | — | |
| ProDepthYear=2024, T.F.=2(-1,0), Data=M, FLOPs.(G)=35.52026.07 | 5.531 | 0.095 | 0.876 | 0.146 | 90.8 | 97.8 | 99.3 | — | — | |
| MTL - FullProtocol=MTL-Full, Backbone=Swin-Tiny [26], Pre-trained=ImageNet-22K [7], Decoder=HRNet2026.03 | 5.76 | — | — | — | — | — | — | -1.03 | 28.49 | |
| FGTOYear=2024, T.F.=1, Data=M+Se, FLOPs.(G)=47.72026.07 | 5.806 | 0.096 | 0.93 | 0.152 | 90.5 | 97.6 | 99.2 | — | — | |
| Flex-Small (Ours)Year=2026, T.F.=1, Data=M, FLOPs.(G)=2.22026.07 | 5.813 | 0.1 | 1.078 | 0.153 | 90.4 | 97.5 | 99.1 | — | — | |
| MTLoRABackbone=Swin-Tiny [26], Pre-trained=ImageNet-22K [7], Decoder=HRNet2026.03 | 5.84 | — | — | — | — | — | — | -1.83 | 7.23 | |
| DynamicDepthTest frames=2 (-1, 0), Semantics=true, W x H=416 x 1282024.07 | 5.867 | 0.103 | 1 | 0.157 | 89.5 | 97.4 | 99.1 | — | — | |
| DynamicDepthYear=2022, T.F.=2(-1,0), Data=M+Se, FLOPs.(G)=18.5+2026.07 | 5.867 | 0.103 | 1 | 0.157 | 89.5 | 97.4 | 99.1 | — | — | |
| PuriLightYear=2026, T.F.=1, Data=M, FLOPs.(G)=5.72026.07 | 5.892 | 0.099 | 1.014 | 0.153 | 90.1 | 97.6 | 99.2 | — | — | |
| Single TaskProtocol=Single Task, Backbone=Swin-Tiny [26], Pre-trained=ImageNet-22K [7], Decoder=HRNet2026.03 | 5.91 | — | — | — | — | — | — | 0 | 56 | |
| Self-DistillationYear=2025, T.F.=1, Data=M, FLOPs.(G)=47.7+2026.07 | 6.076 | 0.118 | 1.091 | 0.173 | 85.4 | 96 | 98.8 | — | — | |
| Flex-Nano (Ours)Year=2026, T.F.=1, Data=M, FLOPs.(G)=0.62026.07 | 6.133 | 0.107 | 1.261 | 0.164 | 89.3 | 97.1 | 98.9 | — | — | |
| ManyDepthTest frames=2 (-1, 0), Semantics=false, W x H=416 x 1282024.07 | 6.223 | 0.114 | 1.193 | 0.17 | 87.5 | 96.7 | 98.9 | — | — | |
| ManyDepthYear=2021, T.F.=2(-1,0), Data=M, FLOPs.(G)=12.12026.07 | 6.223 | 0.114 | 1.193 | 0.17 | 87.5 | 96.7 | 98.9 | — | — | |
| DiTASKBackbone=Swin-Tiny [26], Pre-trained=ImageNet-22K [7], Decoder=HRNet2026.03 | 6.35 | — | — | — | — | — | — | -9.89 | 2.58 | |
| InstaDMTest frames=1, Semantics=false, W x H=832 x 2562024.07 | 6.437 | 0.111 | 1.158 | 0.182 | 86.8 | 96.1 | 98.3 | — | — | |
| MonoVitYear=2022, T.F.=1, Data=M, FLOPs.(G)=47.72026.07 | 6.589 | 0.114 | 1.238 | 0.174 | 86 | 96.5 | 99 | — | — | |
| Lee et al.Test frames=1, Semantics=false, W x H=832 x 2562024.07 | 6.695 | 0.116 | 1.213 | 0.186 | 85.2 | 95.1 | 98.2 | — | — | |
| Monodepth2Test frames=1, Semantics=false, W x H=416 x 1282024.07 | 6.876 | 0.129 | 1.569 | 0.187 | 84.9 | 95.7 | 98.3 | — | — | |
| Videos in the wildTest frames=1, Semantics=false, W x H=416 x 1282024.07 | 6.96 | 0.127 | 1.33 | 0.195 | 83 | 94.7 | 98.1 | — | — | |
| Li et al.Test frames=1, Semantics=false, W x H=416 x 1282024.07 | 6.98 | 0.119 | 1.29 | 0.19 | 84.6 | 95.2 | 98.2 | — | — | |
| Struct2Depth 2Test frames=3 (-1, 0, +1), Semantics=true, W x H=416 x 1282024.07 | 7.024 | 0.151 | 2.492 | 0.202 | 82.6 | 93.7 | 97.2 | — | — | |
| Struct2Depth 2Test frames=1, Semantics=false, W x H=416 x 1282024.07 | 7.28 | 0.145 | 1.737 | 0.205 | 81.3 | 94.2 | 97.6 | — | — | |
| Pilzer et al.Test frames=1, Semantics=false, W x H=512 x 2562024.07 | 8.049 | 0.24 | 4.264 | 0.334 | 71 | 87.1 | 93.7 | — | — | |
| MTL - DecProtocol=MTL-Dec, Backbone=Swin-Tiny [26], Pre-trained=ImageNet-22K [7], Decoder=HRNet2026.03 | 8.54 | — | — | — | — | — | — | -30.59 | 0.97 |