Client is an Europe based manufacturer of network cameras, access control, and network audio devices for the physical security and video surveillance industries. They are well known video surveillance solution provider. The client now wants to upgrade their existing products by adding FPGA hardware capable of running CNN AI model with high speed and ultra-low power with decent FPS that can cater AI/ML solutions. The client approached MosChip to help them build AI based human counting solution which is compatible with FPGA hardware and can be deployed with all the supported CNN layers and able to detect, count humans in given frame. Deploying Convolution Neural Network is challenging task where underlying hardware does not support some of the layers which are generally used in developing computer vision application.