Tracking multiple objects in a video stream using occlusion-aware single-object tracking

A computerized method for tracking multiple objects in a real-time video stream, the computerized method comprising:

• determining, by a processor, a frame type of an individual frame of a plurality of frames of the real-time video stream;
• based on the individual frame being an object detection frame type:
• detecting, by the processor, a set of object proposals in the individual frame;
• calculating affinity scores of each object proposal of the set of object proposals and a current object track of a set of object tracks;
• based on the calculated affinity scores, assigning, by the processor, associations between the set of detected object proposals and the set of object tracks;
• updating, by the processor, statuses of the set of object tracks based on the assigned associations;
• based on the individual frame being an object tracking frame type:
• performing, by the processor, single-object tracking on the individual frame based on each object track of the set of object tracks;
• and
• updating, by the processor, the set of object tracks based on the performed single-object tracking;
• and
• providing, by an output device via the processor, an object location data stream based on the set of object tracks, in real time after processing the individual frame.

One or more computer storage media having computer-executable instructions for tracking multiple objects in a real-time video stream that, upon execution by a processor, cause the processor to at least:

• determine a frame type of an individual frame of a plurality of frames of the real-time video stream;
• based on the individual frame being an object detection frame type:
• detect a set of object proposals in the individual frame;
• calculate affinity scores of each object proposal of the set of object proposals and a current object track of a set of object tracks;
• based on the calculated affinity scores, assign associations between the set of object proposals and the set of object tracks;
• and
• update statuses of the set of object tracks based on the assigned associations;
• based on the individual frame being an object tracking frame type:
• perform single-object tracking on the individual frame based on each object track of the set of object tracks;
• and
• update the set of object tracks based on the performed single-object tracking;
• and
• provide, via an output device, an object location data stream based on the set of object tracks in real time after processing the individual frame.