🌉 Smol (<25M param) models for Real Time Semantic Segmentation
| Model | Params | mIoU | Acc | FPS | Weights |
|---|---|---|---|---|---|
| DeepLabV3+ regnetz_b16 (smp) | 10.7 M | 75.84 | 84.97 | 177.01 | --- |
| DeepLabV3 MobileNetV3 Large (torchvision) | 11.0 M | 68.61 | 79.59 | --- | --- |
| DeepLabV3+ seresnext26d (smp) | 17.9 M | 74.05 | 82.71 | --- | --- |
| UNet seresnext26d (smp) | 23.8 M | 73.41 | 82.34 | --- | --- |
Note
Frames per second (FPS) are measured in an NVIDIA RTX 4090 using float32 precision.