Scene Classification on Places365 (val)
60.14Top-1 AccuracyFOSNet
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FOSNetFusion level=feature, Fusion method=mixed CCM-CCG, Backbone=SE-ResNeXt-101, Evaluation protocol=10-crop testing2019.07 | 60.14 | — | — | 88.86 | — | — | |
| FOSNetFusion level=feature, Fusion method=Concatenate, Backbone=SE-ResNeXt-101, Evaluation protocol=10-crop testing2019.07 | 60.06 | — | — | 88.78 | — | — | |
| FOSNetFusion level=feature, Fusion method=CCG, Backbone=SE-ResNeXt-101, Evaluation protocol=10-crop testing2019.07 | 60.03 | — | — | 88.72 | — | — | |
| FOSNetFusion level=score, Fusion method=CCG, Backbone=SE-ResNeXt-101, Evaluation protocol=10-crop testing2019.07 | 59.92 | — | — | 88.69 | — | — | |
| FOSNetFusion level=feature, Fusion method=Sum, Backbone=SE-ResNeXt-101, Evaluation protocol=10-crop testing2019.07 | 59.82 | — | — | 88.57 | — | — | |
| SE-ResNeXt-101 + SCLTraining loss=Scene Coherence Loss (SCL), Backbone=SE-ResNeXt-101, Evaluation protocol=10-crop testing2019.07 | 59.8 | — | — | 88.66 | — | — | |
| SE-Resnet-152Evaluation protocol=10-crop testing, Backbone=ResNet-1522019.07 | 59.63 | — | — | 88.99 | — | — | |
| SE-ResNeXt-101Implementation source=our implementation, Evaluation protocol=10-crop testing, Backbone=ResNeXt-1012019.07 | 59.1 | — | — | 88.44 | — | — | |
| Places2-365-CNNEvaluation protocol=10-crop testing2019.07 | 58.93 | — | — | 88.52 | — | — | |
| Multi-Resolution CNNsEvaluation protocol=10-crop testing2019.07 | 58.3 | — | — | 87.3 | — | — | |
| SOSF+CFA+GAFEvaluation protocol=10-crop testing2019.07 | 57.27 | — | — | — | — | — | |
| CNN-SMNEvaluation protocol=10-crop testing2019.07 | 57.1 | — | — | — | — | — | |
| CCMEvaluation protocol=10-crop testing2019.07 | 56.82 | — | — | 86.92 | — | — | |
| Semantic-Aware Scene Recognition (Ours)Number of Parameters=~ 47 M2019.09 | 56.51 | — | — | 86 | 71.57 | 56.51 | |
| DenseNet-161Number of Parameters=~ 29 M2019.09 | 56.12 | — | — | 86.12 | 71.48 | 56.12 | |
| ResNet-50Number of Parameters=~ 25 M2019.09 | 55.47 | — | — | 85.36 | 70.4 | 55.47 | |
| Places365-VGGEvaluation protocol=10-crop testing2019.07 | 55.24 | — | — | — | — | — | |
| VGG-19Number of Parameters=~ 143 M, Reported in=[1]2019.09 | 55.24 | — | — | 84.91 | — | — | |
| ResNet-50Number of Parameters=~ 25 M, Reported in=[1]2019.09 | 54.74 | — | — | 85.08 | — | — | |
| GoogLeNetNumber of Parameters=~ 7 M, Reported in=[1]2019.09 | 53.63 | — | — | 83.88 | — | — | |
| AlexNetNumber of Parameters=~ 62 M, Reported in=[1]2019.09 | 53.17 | — | — | 82.89 | — | — | |
| ResNet-18Number of Parameters=~ 12 M2019.09 | 53.05 | — | — | 83.86 | 68.87 | 54.4 | |
| AlexNetNumber of Parameters=~ 62 M2019.09 | 47.45 | — | — | 78.39 | 62.33 | 49.15 | |
| Adi-RedEvaluation protocol=10-crop testing2019.07 | 41.87 | — | — | — | — | — | |
| Semantic BranchNumber of Parameters=~ 2.6 M2019.09 | 36.2 | — | — | 68.48 | 50.11 | 36.2 | |
| B-CNNBackbone=ResNet-50, Prediction=10-crop2019.04 | — | 44.24 | 14.27 | — | — | — | |
| GoogleNetBackbone=GoogleNet, Prediction=10-crop2019.04 | — | 46.37 | 16.12 | — | — | — | |
| iSQRT-COV + E-PNBackbone=ResNet-50, Prediction=10-crop2019.04 | — | 45.34 | 15.13 | — | — | — | |
| iSQRT-COV-NetBackbone=ResNet-50, Prediction=10-crop2019.04 | — | 43.68 | 13.73 | — | — | — | |
| Places-365-CNNevaluation_protocol=single-crop2017.09 | — | 41.07 | 11.48 | — | — | — | |
| ResNet-152evaluation_protocol=single-crop, implementation=ours2017.09 | — | 41.15 | 11.61 | — | — | — | |
| ResNet-152Backbone=ResNet-152, Prediction=10-crop2019.04 | — | 45.26 | 14.92 | — | — | — | |
| ResNet-50Backbone=ResNet-50, Prediction=10-crop2019.04 | — | 44.82 | 14.71 | — | — | — | |
| SE-ResNet-152evaluation_protocol=single-crop2017.09 | — | 40.37 | 11.01 | — | — | — | |
| VGG-VD16Backbone=VGG-VD16, Prediction=10-crop2019.04 | — | 44.76 | 15.09 | — | — | — |