1 Visualizing the test result by generating and saving density maps for each input picture. We normalize the output by its max value for better visualization, which shows each pixel's percentage of crowd count contribution to the whole picture. Different colours means different density of the pixel. We also offer an optional argument "--need-map" for saving density maps.
2 Changing the clerical argument error in README.txt to the right ones.
3 Providing more models such as VGG16 and ResNet18 for more choices.
1 Visualizing the test result by generating and saving density maps for each input picture. We normalize the output by its max value for better visualization, which shows each pixel's percentage of crowd count contribution to the whole picture. Different colours means different density of the pixel. We also offer an optional argument "--need-map" for saving density maps. 2 Changing the clerical argument error in README.txt to the right ones. 3 Providing more models such as VGG16 and ResNet18 for more choices.