Semantic Segmentation on Cityscapes (val) (Pixel-wise ASR and MSE)
82Pixel-wise ASRAdaptive Frequency Trigger
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Adaptive Frequency TriggerQuality=3, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 82 | 7.7 | |
| Adaptive Frequency TriggerQuality=2, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 81 | 11.4 | |
| LIRAQuality=2, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 79.6 | 15.6 | |
| Adaptive Frequency TriggerQuality=1, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 76.4 | 10.8 | |
| Adaptive Frequency TriggerQuality=Mean, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 71.5 | 7.2 | |
| LIRAQuality=3, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 67.7 | 8.4 | |
| Adaptive Frequency TriggerQuality=4, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 66.6 | 5.6 | |
| LIRAQuality=5, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 65.7 | 4.2 | |
| LIRAQuality=4, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 65.6 | 5.7 | |
| Adaptive Frequency TriggerQuality=5, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 64.9 | 4.2 | |
| Adaptive Frequency TriggerQuality=6, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 58.4 | 3.2 | |
| LIRAQuality=Mean, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 56.9 | 7 | |
| LIRAQuality=6, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 56.5 | 2.9 | |
| LIRAQuality=1, Backbone=WideResNet38, Compression Model=Cheng-Anchor2023.02 | 6 | 4.9 |