Semantic Segmentation on NYU v2 (Detailed Class Metrics)
83.26Wall AccuracyHN pre-training
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| HN pre-trainingArch=SegNet, Pre-training=HN-labels2020.02 | 83.26 | 90.63 | 58.81 | 58.4 | 38.79 | 37.05 | 14.24 | 26.16 | 18.24 | 52.92 | |
| ImageNet pre-trainArch=SegNet, Pre-training=ImageNet2020.02 | 81.54 | 86.1 | 58.86 | 37.76 | 35.05 | 38.25 | 4.35 | 25.4 | 17.34 | 50.58 | |
| No pre-trainingArch=SegNet, Pre-training=Scratch2020.02 | 77.57 | 74.91 | 42.36 | 13.25 | 6.58 | 22.42 | 3.17 | 13.73 | 8.78 | 38.74 | |
| CIFAR100 pre-trainArch=SegNet, Pre-training=CIFAR1002020.02 | 66.33 | 66.96 | 19.84 | 0.03 | 0.01 | 0.05 | 0 | 10.7 | 6.23 | 31.9 | |
| HN pre-trainingArch=DeepLab, Pre-training=HN-labels2020.02 | 32 | 77.64 | 87.45 | 60.15 | 55.67 | 60.51 | 53.69 | 56.04 | 33.49 | 62.98 | |
| ImageNet pre-trainArch=DeepLab, Pre-training=ImageNet2020.02 | 28.85 | 74.69 | 87.98 | 64.4 | 56.22 | 59.1 | 50.78 | 57.84 | 34.27 | 61.7 | |
| No pre-trainingArch=DeepLab, Pre-training=Scratch2020.02 | 18.7 | 66.72 | 60.78 | 22.3 | 20.7 | 15.24 | 12.49 | 24.77 | 6.73 | 31.41 |