UniRepLKNet: A Universal Perception Large-Kernel ConvNet for Audio, Video, Point Cloud, Time-Series and Image Recognition

Open paper
Benchmark Model Rank Results
semantic-segmentation-on-ade20kUniRepLKNet-XL#47Validation mIoU: 55.6
semantic-segmentation-on-ade20kUniRepLKNet-L++#53Validation mIoU: 55
semantic-segmentation-on-ade20kUniRepLKNet-B++#74Validation mIoU: 53.9
semantic-segmentation-on-ade20kUniRepLKNet-S++#92Validation mIoU: 52.7
semantic-segmentation-on-ade20kUniRepLKNet-S#115Validation mIoU: 51
semantic-segmentation-on-ade20kUniRepLKNet-T#149Validation mIoU: 49.1
image-classification-on-imagenetUniRepLKNet-XL++#70Top 1 Accuracy: 88%
image-classification-on-imagenetUniRepLKNet-L++#72Top 1 Accuracy: 87.9%
image-classification-on-imagenetUniRepLKNet-B++#101Top 1 Accuracy: 87.4%
image-classification-on-imagenetUniRepLKNet-S++#162Top 1 Accuracy: 86.4%
image-classification-on-imagenetUniRepLKNet-S#409Top 1 Accuracy: 83.9%
image-classification-on-imagenetUniRepLKNet-T#492Top 1 Accuracy: 83.2%
image-classification-on-imagenetUniRepLKNet-N#663Top 1 Accuracy: 81.6%
image-classification-on-imagenetUniRepLKNet-P#754Top 1 Accuracy: 80.2%
image-classification-on-imagenetUniRepLKNet-F#861Top 1 Accuracy: 78.6%
image-classification-on-imagenetUniRepLKNet-A#934Top 1 Accuracy: 77%
object-detection-on-coco-val2017UniRepLKNet-XL (IN-22K, Cascade Mask R-CNN 3x)#85box AP: 56.4Params (M): 443
object-detection-on-coco-val2017UniRepLKNet-L (IN-22K, Cascade Mask R-CNN 3x)#94box AP: 55.8Params (M): 276
object-detection-on-coco-val2017UniRepLKNet-B (IN-22K, Cascade Mask R-CNN 3x)#113box AP: 54.8Params (M): 155
object-detection-on-coco-val2017UniRepLKNet-S (IN-22K, Cascade Mask R-CNN 3x)#126box AP: 54.3Params (M): 113
object-detection-on-coco-val2017UniRepLKNet-S (Cascade Mask R-CNN 3x)#155box AP: 53.0Params (M): 113
object-detection-on-coco-val2017UniRepLKNet-T (Cascade Mask R-CNN 3x)#179box AP: 51.8Params (M): 89

Rank counts only results with a code link.