I implemented an algorithm for pedestrian classification using Keras and Tensorflow and it is quite good, it has 96 percent accuracy on test data. It gives you information about existence of a person on the picture. Now I want to use that to detect where the person is. How to do that? For example, I want to build rectangle around the pedestrian but I don't know how to do that if I already know that there is a person on the picture.
What you are trying to do is called object localisation. If you have multiple of that object in the scene it is called object detection. There are numerous things that you can try that I refer to some of them.
- YOLO paper which may be the perfect thing for your task which has the state of the art performance these days.
- There is another algorithm which is old and is not efficient but is a simple way of detecting objects. It is not that much accurate. It is called Sliding Window.
- The previous algorithm is time-consuming due to the fact that it calculates repetitive calculations. In OverFeat: Integrated Recognition, Localization and ... paper they have tried to overcome that mistake.
There are other approaches too because object detection is an important task in computer vision and every day someone makes another approach. You can see the citation of the mentioned papers which means there are abundant studies about the problem.