Aupera Technologies is a technology innovator which builds intelligent computing and storage systems, with a focus on optimizing video data storage, processing, and analytic applications.
At Aupera Technologies, I was involved in developing advanced applications of Deep Learning in video data analytics, which included:
- Participation in full design cycle in the development of a computer vision software from considering the client's needs to developing the product and optimizing for deployment;
- Researching several different object detection systems and performing machine learning to generate convolutional neural network (CNN) models, capable of both detecting and classifying target objects in images;
- Reducing the size of the CNN models by 75% with negligible loss in performance, in order to meet memory restrictions on embedded devices;
- Providing hardware support by cross-compiling the object detection project into embedded devices and optimizing graph inference computations on ARM CPU, resulting in 50x speed up;
- Building two demos of the project by setting up a server on both an x86 machine and an embedded device, and implementing a web interface.
Unfortunately, I was working under a Non-Disclosure Agreement, so this is as much as I can say about my experience.