RF-Next: Efficient Receptive Field Search for Convolutional Neural Networks

Benchmark Model Rank Results
action-segmentation-on-breakfast-1RF++-SSTDA#23Acc: 70.8
instance-segmentation-on-coco-2017-valRF-ConvNeXt-T Cascade R-CNN#4AP: 44.3
object-detection-on-coco-val2017RF-ConvNeXt-T (Cascade Mask R-CNN)#139box AP: 50.9AP50: 69.5AP75: 55.5APS: 34.3APM: 54.6APL: 65.8
semantic-segmentation-on-imagenet-sRF-ConvNext-Tiny (rfmerge, P4, 224x224, SUP)#8mIoU (val): 51.3mIoU (test): 51.1
semantic-segmentation-on-imagenet-sRF-ConvNext-Tiny (rfmultiple, P4, 224x224, SUP)#9mIoU (val): 50.8mIoU (test): 50.5
semantic-segmentation-on-imagenet-sRF-ConvNext-Tiny (rfsingle, P4, 224x224, SUP)#10mIoU (val): 50.7mIoU (test): 50.5