daimuuc / PyramidScaleNetwork

To the best of our knowledge, this is the first work to explicitly address feature similarity issue in multi-column design. Extensive experiments on four challenging benchmarks (ShanghaiTech, UCF_CC_50, UCF-QNRF, and Mall) demonstrate the effectiveness of the proposed innovations as well as the superior performance over the state-of-the-art. More remarkably, our method can be easily applied to other existing multi-column models as a plug-in to significantly boost the performance to a large extent.

Pyramid_Scale_Network

This is the PyTorch version repo for "Exploit the potential of Multi-column architecture for Crowd Counting", which delivered a state-of-the-art, straightforward and end-to-end architecture for crowd counting tasks. We also recommend another work on crowd counting(Deep Density-aware Count Regressor), which is accepted by ECAI2020.

Datasets

ShanghaiTech Dataset

Prerequisites

We strongly recommend Anaconda as the environment.

Python: 3.6

PyTorch: 1.5.0

Train & Test

1、python make_dataset.py # generate the ground truth. the ShanghaiTech dataset should be placed in the "datasets" directory.
2、python train.py # train model
3、python val.py # test model

Results

partA: MAE 55.5 MSE 90.1

partB: MAE 6.8 MSE 10.7