Open-set classification is a problem of handling 'unknown' classes that are not contained in the training dataset, whereas traditional classifiers assume that only known classes appear in the test environment. Existing open-set classifiers rely on deep networks trained in a supervised manner on known classes in the training set; this causes specialization of learned representations to known classes and makes it hard to distinguish unknowns from knowns. In contrast, we train networks for joint classification and reconstruction of input data. This enhances the learned representation so as to preserve information useful for separating unknowns from knowns, as well as to discriminate classes of knowns. Our novel Classification-Reconstruction learning for Open-Set Recognition (CROSR) utilizes latent representations for reconstruction and enables robust unknown detection without harming the known-class classification accuracy. Extensive experiments reveal that the proposed method outperforms existing deep open-set classifiers in multiple standard datasets and is robust to diverse outliers. The code is available in https://nae-lab.

Abstract-We present an accurate stereo matching method using local expansion moves based on graph cuts. This new move-making scheme is used to efficiently infer per-pixel 3D plane labels on a pairwise Markov random field (MRF) that effectively combines recently proposed slanted patch matching and curvature regularization terms. The local expansion moves are presented as many α-expansions defined for small grid regions. The local expansion moves extend traditional expansion moves by two ways: localization and spatial propagation. By localization, we use different candidate α-labels according to the locations of local α-expansions. By spatial propagation, we design our local α-expansions to propagate currently assigned labels for nearby regions. With this localization and spatial propagation, our method can efficiently infer MRF models with a continuous label space using randomized search. Our method has several advantages over previous approaches that are based on fusion moves or belief propagation; it produces submodular moves deriving a subproblem optimality; it helps find good, smooth, piecewise linear disparity maps; it is suitable for parallelization; it can use cost-volume filtering techniques for accelerating the matching cost computations. Even using a simple pairwise MRF, our method is shown to have best performance in the Middlebury stereo benchmark V2 and V3.

For assistance with grazing cattle management, we propose a cattle detection and counting system based on Convolutional Neural Networks (CNNs) using aerial images taken by an Unmanned Aerial Vehicle (UAV). To improve detection performance, we take advantage of the fact that, with UAV images, the approximate size of the objects can be predicted when the UAV's height from the ground can be assumed to be roughly constant. We resize an image to be fed into the CNN to an optimum resolution determined by the object size and the down-sampling rate of the network, both in training and testing. To avoid repetition of counting in images that have large overlaps to adjacent ones and to obtain the accurate number of cattle in an entire area, we utilize a three-dimensional model reconstructed by the UAV images for merging the detection results of the same target. Experiments show that detection performance is greatly improved when using the optimum input resolution with an F-measure of 0.952, and counting results are close to the ground truths when the movement of cattle is approximately stationary compared to that of the UAV's.

No abstract