Inference Efficiency on 256 x 256 images
0.0058Latency (Sec/Image)ULVM-UNet
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ULVM-UNetResolution=256 x 256, Batch size=1, Hardware=NVIDIA RTX 3090 Ti (24 GB), Timing=CUDA timing, Profiling=PyTorch memory profiling2026.04 | 0.0058 | 17.4 | |
| MambaLiteUNetResolution=256 x 256, Batch size=1, Hardware=NVIDIA RTX 3090 Ti (24 GB), Timing=CUDA timing, Profiling=PyTorch memory profiling2026.04 | 0.0167 | 54.5 | |
| LightM-UNetResolution=256 x 256, Batch size=1, Hardware=NVIDIA RTX 3090 Ti (24 GB), Timing=CUDA timing, Profiling=PyTorch memory profiling2026.04 | 0.0194 | 63.6 | |
| VM-UNetResolution=256 x 256, Batch size=1, Hardware=NVIDIA RTX 3090 Ti (24 GB), Timing=CUDA timing, Profiling=PyTorch memory profiling2026.04 | 0.1718 | 582.5 | |
| VM-UNet2Resolution=256 x 256, Batch size=1, Hardware=NVIDIA RTX 3090 Ti (24 GB), Timing=CUDA timing, Profiling=PyTorch memory profiling2026.04 | 0.1836 | 613.7 |