DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs

Open paper
Benchmark Model Rank Results
fine-grained-image-classification-on-stanfordRDNet-S (224 res, IN-1K pretrained)#37Accuracy: 94.2%FLOPS: 8.7GPARAMS: 50M
fine-grained-image-classification-on-stanfordRDNet-L (224 res, IN-1K pretrained)#38Accuracy: 94.2%FLOPS: 34.7GPARAMS: 186M
fine-grained-image-classification-on-stanfordRDNet-B (224 res, IN-1K pretrained)#39Accuracy: 94.1%FLOPS: 15.4GPARAMS: 87M
fine-grained-image-classification-on-stanfordRDNet-T (224 res, IN-1K pretrained)#43Accuracy: 93.9%FLOPS: 5.0GPARAMS: 24M
image-classification-on-cifar-10RDNet-L (224 res, IN-1K pretrained)#8Percentage correct: 99.31
image-classification-on-cifar-10RDNet-B (224 res, IN-1K pretrained)#9Percentage correct: 99.31
image-classification-on-cifar-10RDNet-T (224 res, IN-1K pretrained)#28Percentage correct: 98.88
image-classification-on-imagenetRDNet-L (384 res)#195Top 1 Accuracy: 85.8%Number of params: 186MGFLOPs: 34.7
image-classification-on-imagenetRDNet-L#286Top 1 Accuracy: 84.8%Number of params: 186MGFLOPs: 34.7
image-classification-on-imagenetRDNet-B#313Top 1 Accuracy: 84.4%Number of params: 87MGFLOPs: 15.4
image-classification-on-imagenetRDNet-S#390Top 1 Accuracy: 83.7%Number of params: 50MGFLOPs: 8.7
image-classification-on-imagenetRDNet-T#489Top 1 Accuracy: 82.8%Number of params: 24MGFLOPs: 5.0
image-classification-on-inaturalist-2018RDNet-L (224 res, IN-1K pretrained)#10Top-1 Accuracy: 81.8%Number of params: 186M
image-classification-on-inaturalist-2018RDNet-B (224 res, IN-1K pretrained)#13Top-1 Accuracy: 80.5Number of params: 87M
image-classification-on-inaturalist-2018RDNet-S (224 res, IN-1K pretrained)#17Top-1 Accuracy: 79.1Number of params: 50M
image-classification-on-inaturalist-2018RDNet-T (224 res, IN-1K pretrained)#21Top-1 Accuracy: 77.0Number of params: 24M
image-classification-on-inaturalist-2019RDNet-L (224 res, IN-1K pretrained)#4Top-1 Accuracy: 83.7Number of params: 186M
image-classification-on-inaturalist-2019RDNet-B (224 res, IN-1K pretrained)#5Top-1 Accuracy: 83.5Number of params: 87M
image-classification-on-inaturalist-2019RDNet-S (224 res, IN-1K pretrained)#6Top-1 Accuracy: 82.9Number of params: 50M
image-classification-on-inaturalist-2019RDNet-T (224 res, IN-1K pretrained)#10Top-1 Accuracy: 81.2Number of params: 24M

Rank counts only results with a code link.