With the mass proliferation of video cameras, many folks are drowning in too much video content, and are unable to make sense of all this data. There's likely not enough hours in the day to watch all these videos and hone in on the few high value segments. One of our customers had a similar problem. This client regularly records an extremely large volume of surgery videos. The video is used to train physicians by having other expert surgeons, health care professionals and crowd workers review the videos and provide strategic feedback. Because surgeries can be very long, they needed a cost effective solution for segmenting the videos so they could more effectively allocate their precious human resources. For example, they wanted to know which segments were in the body, and which were out; they wanted to known when suturing, cautery or other surgical activities occurred. To our knowledge, there were no off the shelf solutions that could work in such a specialized visual domain. In addition, projects like this often take a year or longer to get something into production.
Due to our experience and our Mayetrix platform, we were able to develop, evaluate, and build out a modern deep learning based AI solution and get the models into a production environment with simple, secure REST based API access in just a few weeks. We were then in a great position to regularly deploy new model improvements without any interruption to our customer's business. Following initial model development, we were able to address more challenging AI tasks like actually tracking objects in the scene.