Horizon Line Estimation on HLW (all)
69.97AUC (theta, rho)Places
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PlacesLearning Paradigm=Classification, Loss Function=Softmax, Initialization=Places2016.04 | 69.97 | 67.38 | |
| ImageNetLearning Paradigm=Classification, Loss Function=Softmax, Initialization=ImageNet2016.04 | 69.02 | 67.08 | |
| SalientLearning Paradigm=Classification, Loss Function=Softmax, Initialization=Salient2016.04 | 67.6 | 67.25 | |
| RandomLearning Paradigm=Classification, Loss Function=Softmax, Initialization=Random2016.04 | 67.58 | 62.75 | |
| Best (Regression Huber)Learning Paradigm=Regression, Loss Function=Huber, Initialization=Best Classification Network2016.04 | 67.19 | 67.27 | |
| Best (Regularized Huber)Learning Paradigm=Regression (regularized w/ classification), Loss Function=Huber, Initialization=Best Classification Network2016.04 | 67.18 | 66.66 | |
| Best (Regularized L2)Learning Paradigm=Regression (regularized w/ classification), Loss Function=L2, Initialization=Best Classification Network2016.04 | 63.92 | 64.41 | |
| PoseNetLearning Paradigm=Classification, Loss Function=Softmax, Initialization=PoseNet2016.04 | 61.65 | 63.56 | |
| Best (Regression)Learning Paradigm=Regression, Loss Function=L2, Initialization=Best Classification Network2016.04 | 60.78 | 62.16 | |
| Zhai et al.2016.04 | 58.24 | — | |
| Places (Regression Huber)Learning Paradigm=Regression, Loss Function=Huber, Initialization=Places2016.04 | 57.79 | 58.78 | |
| Lezama et al.2016.04 | 52.59 | — | |
| Places (Regression)Learning Paradigm=Regression, Loss Function=L2, Initialization=Places2016.04 | 46.84 | 49.1 |