Baidu aims to make machine learning more user-friendly. Developing and optimizing relevant algorithms tends to be a task that requires a lot of experience and thorough-going programming skills. Baidu now provides appropriate models online and plans to leverage its cloud services to also take care of training.
The service is called EZDL. The company has first of all released image and sound recognition versions, both of which also enable classification and object recognition. Users can upload their own image and sound files to train the recognition algorithm. According to Baidu, a relatively small number of examples suffice. After completing the training, clients get a ready-made model that they can use either offline by means of SDK or online via cloud API, and which they can install in their own applications. It should, for example, be possible to detect products on an assembly line or detect unusual noises in security areas by means of sound classification