Inference Speed on Monocular Depth Estimation on KITTI raw (test)
3.3Inference Speed (ms) (Titan XP)Lite-Mono-tiny
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Lite-Mono-tinyEncoder Params. (M)=2.0, Encoder FLOPs (G)=2.4, Decoder Params. (M)=0.2, Decoder FLOPs (G)=0.5, Full Model Params. (M)=2.2, Full Model FLOPs (G)=2.9, Input size=640 × 192, Batch size=162022.11 | 3.3 | 12.7 | |
| Monodepth2Encoder Params. (M)=11.2, Encoder FLOPs (G)=4.5, Decoder Params. (M)=3.1, Decoder FLOPs (G)=3.5, Full Model Params. (M)=14.3, Full Model FLOPs (G)=8.0, Input size=640 × 192, Batch size=162022.11 | 3.8 | 14.3 | |
| Lite-Mono-smallEncoder Params. (M)=2.3, Encoder FLOPs (G)=4.1, Decoder Params. (M)=0.2, Decoder FLOPs (G)=0.7, Full Model Params. (M)=2.5, Full Model FLOPs (G)=4.8, Input size=640 × 192, Batch size=162022.11 | 4.3 | 19.2 | |
| Lite-MonoEncoder Params. (M)=2.9, Encoder FLOPs (G)=4.4, Decoder Params. (M)=0.2, Decoder FLOPs (G)=0.7, Full Model Params. (M)=3.1, Full Model FLOPs (G)=5.1, Input size=640 × 192, Batch size=162022.11 | 4.5 | 20 | |
| Lite-Mono-8MEncoder Params. (M)=8.1, Encoder FLOPs (G)=9.5, Decoder Params. (M)=0.6, Decoder FLOPs (G)=1.7, Full Model Params. (M)=8.7, Full Model FLOPs (G)=11.2, Input size=640 × 192, Batch size=162022.11 | 6.5 | 32.2 | |
| R-MSFM3Encoder Params. (M)=0.7, Encoder FLOPs (G)=2.4, Decoder Params. (M)=2.8, Decoder FLOPs (G)=14.1, Full Model Params. (M)=3.5, Full Model FLOPs (G)=16.5, Input size=640 × 192, Batch size=162022.11 | 7.8 | 22.3 | |
| R-MSFM6Encoder Params. (M)=0.7, Encoder FLOPs (G)=2.4, Decoder Params. (M)=3.1, Decoder FLOPs (G)=28.8, Full Model Params. (M)=3.8, Full Model FLOPs (G)=31.2, Input size=640 × 192, Batch size=162022.11 | 13.1 | 41.7 | |
| MonoViT-tinyEncoder Params. (M)=5.6, Encoder FLOPs (G)=7.8, Decoder Params. (M)=4.7, Decoder FLOPs (G)=15.9, Full Model Params. (M)=10.3, Full Model FLOPs (G)=23.7, Input size=640 × 192, Batch size=162022.11 | 13.5 | 47.4 |