Detect Objectionable Content in Live Stream using powerful AI

It was shocking to know about Christchurch Massacre. It was more painful to know that the culprit used Facebook platform for livestreaming of the incident and Facebook was able to know about the incident only after 12 minutes of the incident.

Facebook explains here that it can't control such events from Live Streaming as they don't have a mechanism to know about the content getting live streamed.

 I am not sure where Facebook's R&D team stands in terms of doing Video Analytics on Live Feed and find out what is getting streamed. However, at KazaCam, we do apply AI on Live Feed to identify the objects and do inference using them resulting into face recognition, crowd estimation, Fire and Smoke alarms, detect masked faces, traffic monitoring and many more use cases.

In this blog, it is explained how can we work on Live Feed to identify objects and take corrective actions by inferencing the details. AI Platforms like TensorFlow and Yolo have already trained models which can detect 3000+ kind of different objects. The system can be programmed to scrutinize all live feeds at start and randomly to check black listed objects and take corrective action.