Search

Unobtrusive monitoring of distances between people indoors is a useful tool in the fight against pandemics. A natural resource to accomplish this are surveillance cameras. Unlike previous distance estimation methods, we use a single, overhead, fisheye camera with wide area coverage and propose two approaches. One method leverages a geometric model of the fisheye lens, whereas the other method uses a neural network to predict the 3D-world distance from people-locations in a fisheye image. To evaluate our algorithms, we collected a first-of-its-kind dataset using single fisheye camera, that comprises a wide range of distances between people (1-58 ft) and will be made publicly available. The algorithms achieve 1-2 ft distance error and over 95% accuracy in detecting social-distance violations.

Person re-identification (PRID) has been thoroughly researched in typical surveillance scenarios where various scenes are monitored by side-mounted, rectilinear-lens cameras. To date, few methods have been proposed for fisheye cameras mounted overhead and their performance is lacking. In order to close this performance gap, we propose a multi-feature framework for fisheye PRID where we combine deep-learning, color-based and location-based features by means of novel feature fusion. We evaluate the performance of our framework for various feature combinations on FRIDA, a public fisheye PRID dataset. The results demonstrate that our multi-feature approach outperforms recent appearance-based deep-learning methods by almost 18% points and location-based methods by almost 3% points in matching accuracy. We also demonstrate the potential application of the proposed PRID framework to people counting in large, crowded indoor spaces.

Domain generalization (DG) is a branch of transfer learning that aims to train the learning models on several seen domains and subsequently apply these pre-trained models to other unseen (unknown but related) domains. To deal with challenging settings in DG where both data and label of the unseen domain are not available at training time, the most common approach is to design the classifiers based on the domain-invariant representation features, i.e., the latent representations that are unchanged and transferable between domains. Contrary to popular belief, we show that designing classifiers based on invariant representation features alone is necessary but insufficient in DG. Our analysis indicates the necessity of imposing a constraint on the reconstruction loss induced by representation functions to preserve most of the relevant information about the label in the latent space. More importantly, we point out the trade-off between minimizing the reconstruction loss and achieving domain alignment in DG. Our theoretical results motivate a new DG framework that jointly optimizes the reconstruction loss and the domain discrepancy. Both theoretical and numerical results are provided to justify our approach., Comment: 13 pages, 2 tables

Person re-identification (PRID) from side-mounted rectilinear-lens cameras is a well-studied problem. On the other hand, PRID from overhead fisheye cameras is new and largely unstudied, primarily due to the lack of suitable image datasets. To fill this void, we introduce the "Fisheye Re-IDentification Dataset with Annotations" (FRIDA), with 240k+ bounding-box annotations of people, captured by 3 time-synchronized, ceiling-mounted fisheye cameras in a large indoor space. Due to a field-of-view overlap, PRID in this case differs from a typical PRID problem, which we discuss in depth. We also evaluate the performance of 10 state-of-the-art PRID algorithms on FRIDA. We show that for 6 CNN-based algorithms, training on FRIDA boosts the performance by up to 11.64% points in mAP compared to training on a common rectilinear-camera PRID dataset., Comment: 8 pages

In this paper, we propose a novel domain generalization (DG) framework based on a new upper bound to the risk on the unseen domain. Particularly, our framework proposes to jointly minimize both the covariate-shift as well as the concept-shift between the seen domains for a better performance on the unseen domain. While the proposed approach can be implemented via an arbitrary combination of covariate-alignment and concept-alignment modules, in this work we use well-established approaches for distributional alignment namely, Maximum Mean Discrepancy (MMD) and covariance Alignment (CORAL), and use an Invariant Risk Minimization (IRM)-based approach for concept alignment. Our numerical results show that the proposed methods perform as well as or better than the state-of-the-art for domain generalization on several data sets., Comment: 8 pages, 2 figures, and 1 table. This paper is accepted at 32nd IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2022)

Background subtraction (BGS) is a fundamental video processing task which is a key component of many applications. Deep learning-based supervised algorithms achieve very good perforamnce in BGS, however, most of these algorithms are optimized for either a specific video or a group of videos, and their performance decreases dramatically when applied to unseen videos. Recently, several papers addressed this problem and proposed video-agnostic supervised BGS algorithms. However, nearly all of the data augmentations used in these algorithms are limited to the spatial domain and do not account for temporal variations that naturally occur in video data. In this work, we introduce spatio-temporal data augmentations and apply them to one of the leading video-agnostic BGS algorithms, BSUV-Net. We also introduce a new cross-validation training and evaluation strategy for the CDNet-2014 dataset that makes it possible to fairly and easily compare the performance of various video-agnostic supervised BGS algorithms. Our new model trained using the proposed data augmentations, named BSUV-Net 2.0, significantly outperforms state-of-the-art algorithms evaluated on unseen videos of CDNet-2014. We also evaluate the cross-dataset generalization capacity of BSUV-Net 2.0 by training it solely on CDNet-2014 videos and evaluating its performance on LASIESTA dataset. Overall, BSUV-Net 2.0 provides a ~5% improvement in the F-score over state-of-the-art methods on unseen videos of CDNet-2014 and LASIESTA datasets. Furthermore, we develop a real-time variant of our model, that we call Fast BSUV-Net 2.0, whose performance is close to the state of the art.

Recent methods for people detection in overhead, fisheye images either use radially-aligned bounding boxes to represent people, assuming people always appear along image radius or require significant pre-/post-processing which radically increases computational complexity. In this work, we develop an end-to-end rotation-aware people detection method, named RAPiD, that detects people using arbitrarily-oriented bounding boxes. Our fully-convolutional neural network directly regresses the angle of each bounding box using a periodic loss function, which accounts for angle periodicities. We have also created a new dataset with spatio-temporal annotations of rotated bounding boxes, for people detection as well as other vision tasks in overhead fisheye videos. We show that our simple, yet effective method outperforms state-of-the-art results on three fisheye-image datasets. Code and dataset are available at http://vip.bu.edu/rapid ., Comment: CVPR 2020 OmniCV Workshop paper extended version

Smart buildings use occupancy sensing for various tasks ranging from energy-efficient HVAC and lighting to space-utilization analysis and emergency response. We propose a people counting system which uses a low-resolution thermal sensor. Unlike previous people-counting systems based on thermal sensors, we use an overhead tripwire configuration at entryways to detect and track transient entries or exits. We develop two distinct people counting algorithms for this configuration. To evaluate our algorithms, we have collected and labeled a low-resolution thermal video dataset using the proposed system. The dataset, the first of its kind, is public and available for download. We also propose new evaluation metrics that are more suitable for systems that are subject to drift and jitter.

We present a novel variational generative adversarial network (VGAN) based on Wasserstein loss to learn a latent representation from a face image that is invariant to identity but preserves head-pose information. This facilitates synthesis of a realistic face image with the same head pose as a given input image, but with a different identity. One application of this network is in privacy-sensitive scenarios; after identity replacement in an image, utility, such as head pose, can still be recovered. Extensive experimental validation on synthetic and real human-face image datasets performed under 3 threat scenarios confirms the ability of the proposed network to preserve head pose of the input image, mask the input identity, and synthesize a good-quality realistic face image of a desired identity. We also show that our network can be used to perform pose-preserving identity morphing and identity-preserving pose morphing. The proposed method improves over a recent state-of-the-art method in terms of quantitative metrics as well as synthesized image quality., Comment: 6 pages, 5 figures, 2019 IEEE 29th International Workshop on Machine Learning for Signal Processing (MLSP)

Background subtraction is a basic task in computer vision and video processing often applied as a pre-processing step for object tracking, people recognition, etc. Recently, a number of successful background-subtraction algorithms have been proposed, however nearly all of the top-performing ones are supervised. Crucially, their success relies upon the availability of some annotated frames of the test video during training. Consequently, their performance on completely "unseen" videos is undocumented in the literature. In this work, we propose a new, supervised, background-subtraction algorithm for unseen videos (BSUV-Net) based on a fully-convolutional neural network. The input to our network consists of the current frame and two background frames captured at different time scales along with their semantic segmentation maps. In order to reduce the chance of overfitting, we also introduce a new data-augmentation technique which mitigates the impact of illumination difference between the background frames and the current frame. On the CDNet-2014 dataset, BSUV-Net outperforms state-of-the-art algorithms evaluated on unseen videos in terms of several metrics including F-measure, recall and precision., Comment: 10 pages

Recent studies show that deep neural networks are vulnerable to adversarial examples which can be generated via certain types of transformations. Being robust to a desired family of adversarial attacks is then equivalent to being invariant to a family of transformations. Learning invariant representations then naturally emerges as an important goal to achieve which we explore in this paper within specific application contexts. Specifically, we propose a cyclically-trained adversarial network to learn a mapping from image space to latent representation space and back such that the latent representation is invariant to a specified factor of variation (e.g., identity). The learned mapping assures that the synthesized image is not only realistic, but has the same values for unspecified factors (e.g., pose and illumination) as the original image and a desired value of the specified factor. Unlike disentangled representation learning, which requires two latent spaces, one for specified and another for unspecified factors, invariant representation learning needs only one such space. We encourage invariance to a specified factor by applying adversarial training using a variational autoencoder in the image space as opposed to the latent space. We strengthen this invariance by introducing a cyclic training process (forward and backward cycle). We also propose a new method to evaluate conditional generative networks. It compares how well different factors of variation can be predicted from the synthesized, as opposed to real, images. In quantitative terms, our approach attains state-of-the-art performance in experiments spanning three datasets with factors such as identity, pose, illumination or style. Our method produces sharp, high-quality synthetic images with little visible artefacts compared to previous approaches.

Reliable facial expression recognition plays a critical role in human-machine interactions. However, most of the facial expression analysis methodologies proposed to date pay little or no attention to the protection of a user's privacy. In this paper, we propose a Privacy-Preserving Representation-Learning Variational Generative Adversarial Network (PPRL-VGAN) to learn an image representation that is explicitly disentangled from the identity information. At the same time, this representation is discriminative from the standpoint of facial expression recognition and generative as it allows expression-equivalent face image synthesis. We evaluate the proposed model on two public datasets under various threat scenarios. Quantitative and qualitative results demonstrate that our approach strikes a balance between the preservation of privacy and data utility. We further demonstrate that our model can be effectively applied to other tasks such as expression morphing and image completion.

Deep convolutional neural networks (ConvNets) have been recently shown to attain state-of-the-art performance for action recognition on standard-resolution videos. However, less attention has been paid to recognition performance at extremely low resolutions (eLR) (e.g., 16 x 12 pixels). Reliable action recognition using eLR cameras would address privacy concerns in various application environments such as private homes, hospitals, nursing/rehabilitation facilities, etc. In this paper, we propose a semi-coupled filter-sharing network that leverages high resolution (HR) videos during training in order to assist an eLR ConvNet. We also study methods for fusing spatial and temporal ConvNets customized for eLR videos in order to take advantage of appearance and motion information. Our method outperforms state-of-the-art methods at extremely low resolutions on IXMAS (93.7%) and HMDB (29.2%) datasets.