➜ ~ python
There are several fundamental concepts to be re-emphasized (Here, we took one single concerned object as our example. There might be multiple concerned objects):
- detection: You don’t know whethere there is a concerned object in the field of view or not, which you will know after the detection. And, if there is such a concerned object in the view, the object location is to be given.
- tracking: You know where the concerned object was. Based on the prior knowledge, you are to determine where this object is going to be next?
- location: Both detection and tracking are looked on as locating the concerned object.
- recognition: Only after the concerned object has been located, more detailed information may be recognized afterwards.
Now, let’s test out Deep Object Tracking Implementation in Tensorflow.
➜ tf-adnet-tracking git:(master) ✗ python runner.py by_dataset --vid-path=./data/freeman1/