Towards Component-based Car Detection

Recent results in computer vision have supported the theory that object detectors built in the statistical learning framework can benefit from a two stage learning process, first learning appropriate diagnostic features for the object being trained, and subsequently training an upper-level classifier on the excitation of these part detectors. In this study we develop a hierarchical detection architecture for automobiles. The classifier operates by first locating keypoints in the test image with a well known interest operator. These keypoints are then compared against a corpus of car-specific keypoints learned from the training data. The resulting similarity vector is input into a Support Vector Machine for classification. We compare the performance of our classifier to that of well documented learning algorithms (SVMs and k-Nearest Neighbors) on two separate databases of still images of cars. Our results suggest that part-based detection architectures indeed work well for this class of objects.