Human Pose Estimation on ExLPose Low-light-normal (LL-N)
42.3AP (IoU 0.5:0.95)Ours
Evaluation Results
| Method | Links | |
|---|---|---|
| OursTraining data (LL)=true, Training data (WL)=true, Param. (M)=27.53, Latency (sec)=1.072023.03 | 42.3 | |
| LIME + Baseline-allTraining data (WL)=true, Training data (Enhanced-LL)=true, Param. (M)=27.37, Latency (sec)=1.652023.03 | 38.3 | |
| DA-LLPose (Ours)Architecture=Bottom-up, Training Labels=Well-lit labels alone2024.07 | 35.6 | |
| AdvEntTraining data (LL)=true, Training data (WL)=true, Param. (M)=27.37, Latency (sec)=1.072023.03 | 35.6 | |
| OursDetection Model=Cascade R-CNN, Components=LSBN + LUPI2023.03 | 35.6 | |
| LLFlow + Baseline-allTraining data (WL)=true, Training data (Enhanced-LL)=true, Param. (M)=66.23, Latency (sec)=3.342023.03 | 35.2 | |
| DANNTraining data (LL)=true, Training data (WL)=true, Param. (M)=27.37, Latency (sec)=1.072023.03 | 34.9 | |
| Baseline-allTraining data (LL)=true, Training data (WL)=true, Param. (M)=27.37, Latency (sec)=1.072023.03 | 33.8 | |
| Baseline-lowTraining data (LL)=true, Param. (M)=27.37, Latency (sec)=1.072023.03 | 32.6 | |
| LIMEArchitecture=Top-down2024.07 | 31.9 | |
| LIME + Baseline-allDetection Model=Cascade R-CNN, Pre-processing=LIME2023.03 | 31.9 | |
| LLFlow†Architecture=Top-down, Training Labels=Low-light and well-lit2024.07 | 28.7 | |
| LLFlow + Baseline-allDetection Model=Cascade R-CNN, Pre-processing=LLFlow2023.03 | 28.7 | |
| DANNArchitecture=Top-down2024.07 | 28 | |
| DANNDetection Model=Cascade R-CNN, Method Category=UDA2023.03 | 28 | |
| CPN-all†Architecture=Top-down, Training Labels=Low-light and well-lit2024.07 | 26.4 | |
| Baseline-allDetection Model=Cascade R-CNN, Training Data=All2023.03 | 26.4 | |
| CPN-lowArchitecture=Top-down2024.07 | 25.4 | |
| Baseline-lowDetection Model=Cascade R-CNN, Training Data=Low-light2023.03 | 25.4 | |
| Baseline-wellTraining data (WL)=true, Param. (M)=27.37, Latency (sec)=1.072023.03 | 23.5 | |
| CPN-well†Architecture=Top-down, Training Labels=Low-light and well-lit2024.07 | 15.3 | |
| Baseline-wellDetection Model=Cascade R-CNN, Training Data=Well-lit2023.03 | 15.3 |