README.md

Counting_With_Adaptive_Auxiliary

Object Counting with Adaptive Auxiliary Tasks

Data Prepare

SHA

• Download checkpoint_best.pth and put it into ./checkpoints/

prediction

python test_SHA.py  

QNRF

• TODO

JHU-Crowd

• TODO

NWPU-Crowd

Our method achieved 76.4 MAE and 327.4 MSE on NWPU-Crowd counting benchmark, which is currently the largest crowd counting benckmark