C++ project, which use Tensorflow DLL and use Opencv DLL to do predict and train.

This DEMO project runs a vision task on Video from Elevator Security Camera. And to count the head and get elevator's running information.

For head counting, this project use a pre-trained frozen PB file which use the Keras-Faster-RCNN network, and head rect data from MPII dataset. For floor and arrow information, this project use a semi-manually data and a simple 2-layer Nerual Network trained frozen PB file.

• tensorflow.dll link:https://pan.baidu.com/s/1jIBriPw pwd:b4ay

• tensorflow.lib link: https://pan.baidu.com/s/1eS2BSQ2 pwd: t2a9

• libprotobuf.lib link: https://pan.baidu.com/s/1kVMK3Uv pwd: kbfv

Task List

• feed network with two/more inputs (NO UI)

• get two/more output from predict (NO UI)

• image process: get a specified rect from image to a Tensorflow Tensor

• load RPN network and predict (NO UI)

• implement helper to get ROI as input for classifier network input (NO UI)