CODA

Implementation for CODA: Counting Objects via Scale-aware Adversarial Adaption

* data
       * dataset folders

* exp
       * experiments logs

* model
       * %dataset name%.yml: dataset configurations
       * %dataset name%_adv.yml: dataset configurations for adversarial training and testing

* src/lib
       * dataset
            * DataPrepare.py: preparing the train/val/test list for training/testing
            * mall/shanghaitech/trancos.py: dataset redefinition
       * network
            * cn.py: counting network definition
            * discriminator.py: discriminator definition
       * opt
            * train.py: training operation for counting network
            * train_adv.py: training operation for adversarial training
            * lr_policy.py: lr policy 
       * utils
            * image_opt.py: image visualization
            * Logger.py: tensorboard logger

* tools
       * demo.py
       * demo_adv.py
       * train_net.py
       * train_net_adv.py

pretrain CN

python tools/train_net.py

density adaption

python tools/train_net_adv.py