A comprehensive content filtering platform with minimal human supervision classifies input stream fast and accurately. It does not depend on listing or text surrounding the content or metadata. It analyzes content itself. Depending on business needs, it can do it in real time or asynchronously offering different hardware budgets.
Hybrid algorithms for deep learning and image fingerprinting provide a computationally efficient method for video/image classification. Deep learning subsystem was pre-trained on 500+ hours of adult content video selected with the goal of creating representative sample.
Integrate an image recognition service that identifies a replacement part in an image provided by the customer and refers him to the appropriate product in the retailer’s online store.
A deep convolutional network is trained on the images provided by the retailer. API for the retailer’s mobile app is provided. The customer takes photos of unknown parts. A part is identified either directly inside a smartphone or on a server depending on the size of the catalogue. The customer is directed to the appropriate item in the catalogue.