Pairwise Decomposition of Image Sequences for Active Multi-View Recognition
It addresses the problem of robust object recognition from arbitrary camera trajectories for applications like robotics or surveillance, offering a novel deep learning approach that is more flexible than traditional geometric methods.
The paper tackles multi-view object recognition by decomposing image sequences into pairs, classifying each with CNNs and weighting contributions, and extends this to active recognition by predicting the next-best-view and optimizing trajectories. It achieves state-of-the-art results on the ModelNet dataset, handling various image types like depth and greyscale.
A multi-view image sequence provides a much richer capacity for object recognition than from a single image. However, most existing solutions to multi-view recognition typically adopt hand-crafted, model-based geometric methods, which do not readily embrace recent trends in deep learning. We propose to bring Convolutional Neural Networks to generic multi-view recognition, by decomposing an image sequence into a set of image pairs, classifying each pair independently, and then learning an object classifier by weighting the contribution of each pair. This allows for recognition over arbitrary camera trajectories, without requiring explicit training over the potentially infinite number of camera paths and lengths. Building these pairwise relationships then naturally extends to the next-best-view problem in an active recognition framework. To achieve this, we train a second Convolutional Neural Network to map directly from an observed image to next viewpoint. Finally, we incorporate this into a trajectory optimisation task, whereby the best recognition confidence is sought for a given trajectory length. We present state-of-the-art results in both guided and unguided multi-view recognition on the ModelNet dataset, and show how our method can be used with depth images, greyscale images, or both.