build details

Show: section status errors & todos local changes recent changes last change in-page changes feedback controls

Object classification


Modified 2018-12-20 by ι™ˆδΊ‘

Nick and David


Object classification is the critical task in comoputer vision applications. It is the task of classifying objects from different object categories. It is useful for duckiebot to classify the objects in the recieved images and it can be helpful in tasks such as object detection and tracking. In duckietown, there are many informative objects such as traffic signs, duckiebots, duckies, house models, etc.

Traditional ways of object classification extract features or descriptors for different classes first and then use classifiers to classify objects. Typical features used are Histogram of Oriented Gradients (HOG), Scale-Invariant Feature Transform (SIFT), Haar-like features, Speeded Up Robust Feature (SURF) etc. Nowadays, with the development of CNNs, the accuracy of object classification improved astoundingly.

An object classification algorithm takes images as input and output what objects it contains. For simplification, HOG plus two-class (binary) classifiers are explained because it has the best performance in traditional algorithms.

No questions found. You can ask a question on the website.