Augmenting Convolutional networks with attention-based aggregation

Open paper
Benchmark Model Rank Results
semantic-segmentation-on-ade20k-valPatchConvNet-L120 (UperNet)#40mIoU: 52.9
semantic-segmentation-on-ade20k-valPatchConvNet-B120 (UperNet)#42mIoU: 52.8
semantic-segmentation-on-ade20k-valPatchConvNet-B60 (UperNet)#46mIoU: 51.1
semantic-segmentation-on-ade20k-valPatchConvNet-S60 (UperNet)#57mIoU: 49.3
semantic-segmentation-on-ade20kPatchConvNet-L120 (UperNet)#88Validation mIoU: 52.9
semantic-segmentation-on-ade20kPatchConvNet-B120 (UperNet)#90Validation mIoU: 52.8
semantic-segmentation-on-ade20kPatchConvNet-B60 (UperNet)#111Validation mIoU: 51.1
semantic-segmentation-on-ade20kPatchConvNet-S60 (UperNet)#145Validation mIoU: 49.3
image-classification-on-imagenetPatchConvNet-L120-21k-384#113Top 1 Accuracy: 87.1%Number of params: 334.3M
image-classification-on-imagenetPatchConvNet-B60-21k-384#147Top 1 Accuracy: 86.5%Number of params: 99.4M
image-classification-on-imagenetPatchConvNet-S60-21k-512#245Top 1 Accuracy: 85.4%Number of params: 25.2M
image-classification-on-imagenetPatchConvNet-B120#379Top 1 Accuracy: 84.1%Number of params: 188.6M
image-classification-on-imagenetPatchConvNet-B60#451Top 1 Accuracy: 83.5%Number of params: 99.4M
image-classification-on-imagenetPatchConvNet-S120#487Top 1 Accuracy: 83.2%Number of params: 47.7M
image-classification-on-imagenetPatchConvNet-S60#610Top 1 Accuracy: 82.1%Number of params: 25.2M
object-detection-on-coco-val2017PatchConvNet-S120 (Mask R-CNN)#282box AP: 47.0
object-detection-on-coco-val2017PatchConvNet-S60 (Mask R-CNN)#290box AP: 46.4

Rank counts only results with a code link.