In this paper, the authors address the problem of automatic object classification from images. Specifically, the authors develop three approaches: (1) Haar-like feature based combined with Cascaded Adaboost Classifier
(2) Histogram of Oriented Gradient combined with Support Vector Machine and (3) Gist descriptor using K- Nearest Neighbors. Currently, each method has been shown to be efficient for specific kinds of objects (1 - hand posture, 2 - standing person, 3 - natural scene). the authors then evaluate the performance (in term of computational time and precision) of each method on a predefined and challenge database containing 10 types of object classes. The experimental results allow us to choose the most robust one to deploy an image content based ads system.