I am using the tensorflow object detection api on windows system and it has been working fine. I am using the pre-trained model as of now which is ssd_mobilenet_v1_coco_11_06_2017. It is easily detecting all the objects in a given input video. I wanted to record time of each object so lets say, if in a video a car enters at 00:05 and leaves at 00:15 which means that it was in for 10secs.
To achieve this, I was looking if there is anything like id of each detected object which the API returns so that I can start a timer from the code to calculate the time of an object. Is there any already built functionality for this in the API.?
Tensorflow Object detection does not provide such functionality, but you can user KFC algorithm(easily available using Open CV) to track the object.
https://www.docs.opencv.org/3.4.1/d2/dff/classcv_1_1TrackerKCF.html
or You can implement SORT above object detection API which uses Kalman Filter but easy to integrate.