Rethinking Spatial Invariance of Convolutional Networks for Object Counting

Benchmark Model Rank Results
crowd-counting-on-jhu-crowdGauNet (ResNet-50)#2MAE: 58.2MSE: 245.1
crowd-counting-on-shanghaitech-aGauNet (ResNet-50)#6MAE: 54.8MSE: 89.1
crowd-counting-on-shanghaitech-bGauNet (ResNet-50)#3MAE: 6.0
crowd-counting-on-ucf-cc-50GauNet (ResNet-50)#3MAE: 186.3
crowd-counting-on-ucf-qnrfGauNet (ResNet-50)#9MAE: 81.6