Abstract:Compressed sensing (CS) has emerged to overcome the inefficiency of Nyquist sampling. However, traditional optimization-based reconstruction is slow and can not yield an exact image in practice. Deep learning-based reconstruction has been a promising alternative to optimization-based reconstruction, outperforming it in accuracy and computation speed. Finding an efficient sampling method with deep learning-based reconstruction, especially for Fourier CS remains a challenge. Existing joint optimization of sampling-reconstruction works ($\mathcal{H}_1$) optimize the sampling mask but have low potential as it is not adaptive to each data point. Adaptive sampling ($\mathcal{H}_2$) has also disadvantages of difficult optimization and Pareto sub-optimality. Here, we propose a novel adaptive selection of sampling-reconstruction ($\mathcal{H}_{1.5}$) framework that selects the best sampling mask and reconstruction network for each input data. We provide theorems that our method has a higher potential than $\mathcal{H}_1$ and effectively solves the Pareto sub-optimality problem in sampling-reconstruction by using separate reconstruction networks for different sampling masks. To select the best sampling mask, we propose to quantify the high-frequency Bayesian uncertainty of the input, using a super-resolution space generation model. Our method outperforms joint optimization of sampling-reconstruction ($\mathcal{H}_1$) and adaptive sampling ($\mathcal{H}_2$) by achieving significant improvements on several Fourier CS problems.
Abstract:Anomaly detection and localization are essential in many areas, where collecting enough anomalous samples for training is almost impossible. To overcome this difficulty, many existing methods use a pre-trained network to encode input images and non-parametric modeling to estimate the encoded feature distribution. In the modeling process, however, they overlook that position and neighborhood information affect the distribution of normal features. To use the information, in this paper, the normal distribution is estimated with conditional probability given neighborhood features, which is modeled with a multi-layer perceptron network. At the same time, positional information can be used by building a histogram of representative features at each position. While existing methods simply resize the anomaly map into the resolution of an input image, the proposed method uses an additional refine network that is trained from synthetic anomaly images to perform better interpolation considering the shape and edge of the input image. For the popular industrial dataset, MVTec AD benchmark, the experimental results show \textbf{99.52\%} and \textbf{98.91\%} AUROC scores in anomaly detection and localization, which is state-of-the-art performance.
Abstract:We introduce N-ImageNet, a large-scale dataset targeted for robust, fine-grained object recognition with event cameras. The dataset is collected using programmable hardware in which an event camera consistently moves around a monitor displaying images from ImageNet. N-ImageNet serves as a challenging benchmark for event-based object recognition, due to its large number of classes and samples. We empirically show that pretraining on N-ImageNet improves the performance of event-based classifiers and helps them learn with few labeled data. In addition, we present several variants of N-ImageNet to test the robustness of event-based classifiers under diverse camera trajectories and severe lighting conditions, and propose a novel event representation to alleviate the performance degradation. To the best of our knowledge, we are the first to quantitatively investigate the consequences caused by various environmental conditions on event-based object recognition algorithms. N-ImageNet and its variants are expected to guide practical implementations for deploying event-based object recognition algorithms in the real world.