Multiclass object recognition inspired by the ventral visual pathway

• Main Author: Mutch, James Vincent
• Language: English
• Published: 2010
• Online Access: http://hdl.handle.net/2429/18090

We describe a biologically-inspired system for classifying objects in still images. Our system learns to identify the class (car, person, etc.) of a previously-unseen instance of an object. As the primate visual system still outperforms computer vision systems on this task by a wide margin, we base our work on a model of the ventral visual pathway, thought to be primarily responsible for object recognition in cortex. Our model modifies that of Serre, Wolf, and Poggio, which hierarchically builds up feature selectivity and invariance to position and scale in a manner analogous to that of visual areas V1, V2, V4, and IT. As in that work, we first apply Gabor filters at all positions and scales; selectivity and invariance are then built up by alternating template matching and max pooling operations. We refine the approach in several biologically plausible ways, using simple versions of sparsification and lateral inhibition. We demonstrate the value of retaining some position and scale information above the intermediate feature level. Using feature selection we arrive at a model that performs better with fewer features. Our final model is tested on the Caltech 101 object categories and the UIUC car localization task, in both cases achieving state-of-the-art performance. The results strengthen the case for using this type of model in computer vision. === Science, Faculty of === Computer Science, Department of === Graduate