Challenges in object detection
In the past, several approaches for object detection were proposed. However, these either perform well in a controlled environment or look for special objects in images like a human face. Even in the case of faces, the approaches suffer from issues like low light conditions, a highly occluded face or tiny face size compared to the image size.
Following are several challenges that are faced by an object detector in real-world applications:
Occlusion: Objects like dogs or cats can be hidden behind one another, as a result, the features that can be extracted from them are not strong enough to say that they are an object.
Viewpoint changes: In cases of different viewpoints of an object, the shape may change drastically and hence the features of the object will also change drastically. This causes a detector which is trained to see a given object from one viewpoint to fail on seeing it from other viewpoints. For example, in the case of person detection, if the detector...
