Panoptic Segmentation on COCO 2014 (val)
46.7PQMOAT-4
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MOAT-4Backbone=Panoptic-DeepLab, params=512.0M, FLOPs=1134.7B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-22K2022.10 | 46.7 | 49.5 | 42.4 | |
| MOAT-3Backbone=Panoptic-DeepLab, params=208.3M, FLOPs=493.3B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-22K2022.10 | 45.4 | 48.3 | 41.1 | |
| SWideRNetBackbone=Panoptic-DeepLab, params=752.5M, FLOPs=2614.0B, Input size=641x641, Inference scale=single-scale2022.10 | 44.4 | — | — | |
| MOAT-2Backbone=Panoptic-DeepLab, params=88.5M, FLOPs=199.7B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-22K2022.10 | 43.9 | 45.9 | 40.8 | |
| ConvNeXt-XLBackbone=Panoptic-DeepLab, params=379.6M, FLOPs=544.1B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-22K2022.10 | 43 | 44.9 | 40 | |
| MOAT-1Backbone=Panoptic-DeepLab, params=53.1M, FLOPs=119.7B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-1K2022.10 | 43 | 44.7 | 40.4 | |
| ConvNeXt-LBackbone=Panoptic-DeepLab, params=220.1M, FLOPs=312.8B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-22K2022.10 | 41.9 | 43.6 | 39.4 | |
| ConvNeXt-BBackbone=Panoptic-DeepLab, params=103.8M, FLOPs=146.2B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-22K2022.10 | 41.7 | 43.6 | 38.9 | |
| MOAT-0Backbone=Panoptic-DeepLab, params=39.5M, FLOPs=76.8B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-1K2022.10 | 41 | 42.6 | 38.6 | |
| ConvNeXt-SBackbone=Panoptic-DeepLab, params=61.9M, FLOPs=87.2B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-1K2022.10 | 40 | 41.4 | 37.9 | |
| Xception-71Backbone=Panoptic-DeepLab, FLOPs=109.2B, Input size=641x641, Inference scale=single-scale2022.10 | 38.9 | — | — | |
| ConvNeXt-TBackbone=Panoptic-DeepLab, params=40.3M, FLOPs=51.3B, Input size=641x641, Inference scale=single-scale, Pre-training=ImageNet-1K2022.10 | 36.7 | 37.3 | 35.7 | |
| ResNet50Backbone=Panoptic-DeepLab, FLOPs=77.8B, Input size=641x641, Inference scale=single-scale2022.10 | 35.1 | — | — | |
| MobileNet-V3Backbone=Panoptic-DeepLab, FLOPs=12.2B, Input size=641x641, Inference scale=single-scale2022.10 | 30 | — | — |