Convolutional Transformer based Dual Discriminator Generative Adversarial Networks for Video Anomaly Detection

Anomaly detection in videos is the task of identifying frames from a video sequence that depict events that do not conform to expected behavior, which is an extremely challenging task due to the ambiguous and unbounded properties of anomalies. With the development of deep learning, video anomaly detection methods based on deep neural networks have made great progress. The existing methods mainly follow two routes, namely, frame reconstruction and frame prediction. Due to the powerful generalization ability of neural networks, the application of reconstruction-based methods is limited. Recently, anomaly detection methods based on prediction have achieved advanced performance. However, their performance suffers when they cannot guarantee lower prediction errors for normal events. In this paper, we propose a novel future frame prediction model based on a bidirectional retrospective generation adversarial network (BR-GAN) for anomaly detection. To predict a future frame with higher quality for normal events, first, we propose a bidirectional prediction combined with a retrospective prediction method to fully mine the bidirectional temporal information between the predicted frame and the input frame sequence. Then, the intensity and gradient loss between the predicted frame and the actual frame together with an adversarial loss are used for appearance (spatial) constraints. In addition, we propose a sequence discriminator composed of a 3-dimensional (3D) convolutional neural network to capture the long-term temporal relationships between frame sequences composed of predicted frames and input frames; this network plays a crucial role in maintaining the motion (temporal) consistency of the predicted frames for normal events. Such appearance and motion constraints further facilitate future frame prediction for normal events, and thus, the prediction network can be highly capable of distinguishing normal and abnormal patterns. Extensive experiments on benchmark datasets demonstrate that our method outperforms most existing state-of-the-art methods, validating the effectiveness of our method for anomaly detection.

Video anomaly detection is of great significance due to its wide applications in video surveillance. Recently, there is a trend of using a video prediction framework to tackle this problem. This kind of methods detects anomalies according to the difference between a predicted frame and its ground truth. However, existing prediction methods lack the consideration of multi-scale temporal constraints on generating future frames. Therefore, based on the prediction framework, this paper proposes a temporal enhanced anomaly detection approach, which designs a generative adversarial network with dual discriminator (frame discriminator and sequence discriminator) to predict future frames. In order to obtain more realistic predictions, other than commonly used spatial constraints and the adversarial penalty from the frame discriminator, we also consider both short-range and long-range motions to impose constraints. Specifically, for short-range motion modeling, we utilize the optical flow loss to ensure temporal continuity over two adjacent frames, while for long-range motion modeling, we design a sequence discriminator to identify fake contained sequences from real sequences, making frames more consistent with their previous consecutive frames. Experiments on three datasets, UCSD Ped1, UCSD Ped2 and Avenue, demonstrate the effectiveness of our method in terms of various evaluation criteria for video anomaly detection.

Video anomaly detection aims to find the events in a video that do not conform to the expected behavior. The prevalent methods mainly detect anomalies by snippet reconstruction or future frame prediction error. However, the error is highly dependent on the local context of the current snippet and lacks the understanding of normality. To address this issue, we propose to detect anomalous events not only by the local context, but also according to the consistency between the testing event and the knowledge about normality from the training data. Concretely, we propose a novel two-stream framework based on context recovery and knowledge retrieval, where the two streams can complement each other. For the context recovery stream, we propose a spatiotemporal U-Net which can fully utilize the motion information to predict the future frame. Furthermore, we propose a maximum local error mechanism to alleviate the problem of large recovery errors caused by complex foreground objects. For the knowledge retrieval stream, we propose an improved learnable locality-sensitive hashing, which optimizes hash functions via a Siamese network and a mutual difference loss. The knowledge about normality is encoded and stored in hash tables, and the distance between the testing event and the knowledge representation is used to reveal the probability of anomaly. Finally, we fuse the anomaly scores from the two streams to detect anomalies. Extensive experiments demonstrate the effectiveness and complementarity of the two streams, whereby the proposed two-stream framework achieves state-of-the-art performance on ShanghaiTech, Avenue and Corridor datasets among the methods without object detection. Even if compared with the methods using object detection, our method reaches competitive or better performance on the ShanghaiTech, Avenue, and Ped2 datasets.

Video anomaly detection (VAD) remains a challenging task in the pattern recognition community due to the ambiguity and diversity of abnormal events. Existing deep learning-based VAD methods usually leverage proxy tasks to learn the normal patterns and discriminate the instances that deviate from such patterns as abnormal. However, most of them do not take full advantage of spatial-temporal correlations among video frames, which is critical for understanding normal patterns. In this paper, we address unsupervised VAD by learning the evolution regularity of appearance and motion in the long and short-term and exploit the spatial-temporal correlations among consecutive frames in normal videos more adequately. Specifically, we proposed to utilize the spatiotemporal long short-term memory (ST-LSTM) to extract and memorize spatial appearances and temporal variations in a unified memory cell. In addition, inspired by the generative adversarial network, we introduce a discriminator to perform adversarial learning with the ST-LSTM to enhance the learning capability. Experimental results on standard benchmarks demonstrate the effectiveness of spatial-temporal correlations for unsupervised VAD. Our method achieves competitive performance compared to the state-of-the-art methods with AUCs of 96.7%, 87.8%, and 73.1% on the UCSD Ped2, CUHK Avenue, and ShanghaiTech, respectively.

Training-free video anomaly detection (VAD) has recently emerged as a scalable alternative to supervised approaches, yet existing methods largely rely on static prompting and geometry-agnostic feature fusion. As a result, anomaly inference is often reduced to shallow similarity matching over Euclidean embeddings, leading to unstable predictions and limited interpretability, especially in complex or hierarchically structured scenes. We introduce MM-VAD, a geometry-aware semantic reasoning framework for training free VAD that reframes anomaly detection as adaptive test-time inference rather than fixed feature comparison. Our approach projects caption-derived scene representations into hyperbolic space to better preserve hierarchical structure and performs anomaly assessment through an adaptive question answering process over a frozen large language model. A lightweight, learnable prompt is optimised at test time using an unsupervised confidence-sparsity objective, enabling context-specific calibration without updating any backbone parameters. To further ground semantic predictions in visual evidence, we incorporate a covariance-aware Mahalanobis refinement that stabilises cross-modal alignment. Across four benchmarks, MM-VAD consistently improves over prior training-free methods, achieving 90.03% AUC on XD-Violence and 83.24%, 96.95%, and 98.81% on UCF-Crime, ShanghaiTech, and UCSD Ped2, respectively. Our results demonstrate that geometry-aware representation and adaptive semantic calibration provide a principled and effective alternative to static Euclidean matching in training-free VAD.

Due to the limited availability of anomaly examples, video anomaly detection is often seen as one-class classification (OCC) problem. A popular way to tackle this problem is by utilizing an autoencoder (AE) trained only on normal data. At test time, the AE is then expected to reconstruct the normal input well while reconstructing the anomalies poorly. However, several studies show that, even with normal data only training, AEs can often start reconstructing anomalies as well which depletes their anomaly detection performance. To mitigate this, we propose a temporal pseudo anomaly synthesizer that generates fake-anomalies using only normal data. An AE is then trained to maximize the reconstruction loss on pseudo anomalies while minimizing this loss on normal data. This way, the AE is encouraged to produce distinguishable reconstructions for normal and anomalous frames. Extensive experiments and analysis on three challenging video anomaly datasets demonstrate the effectiveness of our approach to improve the basic AEs in achieving superiority against several existing state-of-the-art models.

We propose a novel unsupervised approach based on a two-stage object-centric adversarial framework that only needs object regions for detecting frame-level local anomalies in videos. The first stage consists in learning the correspondence between the current appearance and past gradient images of objects in scenes deemed normal, allowing us to either generate the past gradient from current appearance or the reverse. The second stage extracts the partial reconstruction errors between real and generated images (appearance and past gradient) with normal object behaviour, and trains a discriminator in an adversarial fashion. In inference mode, we employ the trained image generators with the adversarially learned binary classifier for outputting region-level anomaly detection scores. We tested our method on four public benchmarks, UMN, UCSD, Avenue and ShanghaiTech and our proposed object-centric adversarial approach yields competitive or even superior results compared to state-of-the-art methods.

Video anomaly detection is an essential task in computer vision which attracts massive attention from academia and industry. The existing approaches are implemented in diverse deep learning frameworks and settings, making it difficult to reproduce the results published by the original authors. Undoubtedly, this phenomenon is detrimental to the development of Video Anomaly detection and community communication. In this paper, we present a PyTorch-based video anomaly detection toolbox, namely PyAnomaly that contains high modular and extensible components, comprehensive and impartial evaluation platforms, a friendly manageable system configuration, and the abundant engineering deployment functions. To make it easy-to-use and easy-to-extend, we implement the architecture by hooks and registers functionality. Remarkably, we have reproduced the comparable experimental results of six representative methods as those published by the original authors, and we will release these pre-trained models with more rich configurations. To our best knowledge, the PyAnomaly is the first open-source tool in video anomaly detection and is available at https://github.com/YuhaoCheng/PyAnomaly.

Future frame prediction is a challenging task in computer vision with practical applications in areas such as video generation, autonomous driving, and robotics. Traditional recurrent neural networks have limited effectiveness in capturing long-range dependencies between frames, and combining convolutional neural networks (CNNs) with recurrent networks has limitations in modeling complex dependencies. Generative adversarial networks have shown promising results, but they are computationally expensive and suffer from instability during training. In this article, we propose a novel approach for future frame prediction that combines the encoding capabilities of 3-D CNNs with the sequence modeling capabilities of Transformers. We also propose a spatial self-attention mechanism and a novel neighborhood pixel intensity loss to preserve structural information and local intensity, respectively. Our approach outperforms existing methods in terms of structural similarity (SSIM), peak signal-to-noise ratio (PSNR), and learned perceptual image patch similarity (LPIPS) scores on five public datasets. More precisely, our model exhibited an average improvement of 4.64%, 18.5%, and 42% concerning SSIM, PSNR, and LPIPS for the second most proficient method correspondingly, across all datasets. The results demonstrate the effectiveness of our proposed method in generating high-quality predictions of future frames.

Meteorological imagery prediction is an important and challenging problem for weather forecasting. It can also be seen as a video frame prediction problem that estimates future frames based on observed meteorological imageries. Despite it is a widely-investigated problem, it is still far from being solved. Current state-of-the-art deep learning based approaches mainly optimise the mean square error loss resulting in blurry predictions. We address this problem by introducing a Meteorological Predictive Learning GAN model (in short MPL-GAN) that utilises the conditional GAN along with the predictive learning module in order to handle the uncertainty in future frame prediction. Experiments on a real-world dataset demonstrate the superior performance of our proposed model. Our proposed model is able to map the blurry predictions produced by traditional mean square error loss based predictive learning methods back to their original data distributions, hence it is able to improve and sharpen the prediction. In particular, our MPL-GAN achieves an average sharpness of 52.82, which is 14% better than the baseline method. Furthermore, our model correctly detects the meteorological movement patterns that traditional unconditional GANs fail to do.

Anomaly detection in videos is a long-standing and challenging problem. Previous methods often adopt deep and large neural networks to achieve the best detection accuracy; however, the high computational costs prevent them from being used in real-world applications with constrained computational resources. In this paper, we develop a mem ory- a ugmented tr ansformer named MemATr, which is capable of detecting video anomalies effectively. The proposed network is lightweight and can be easily deployed on mobile devices. Furthermore, we propose a memory transformer module to make predictions that are closer to normal inputs, thereby leading to a higher error for abnormal input patterns. Memory-attention is the main component of the proposed memory transformer, which can retrieve the features from learnable values rather than from the backbone like previous methods. Extensive experiments on the UCSD Ped2, CUHK Avenue, and ShanghaiTech benchmarks can demonstrate that our model has a significantly smaller model size while still achieving competitive detection accuracy. Our model has only 1/12 the number of parameters of the baseline model. Besides, our model achieves a 4.6% increase in accuracy on the ShanghaiTech dataset and has roughly the same accuracy compared with the baseline on the other two datasets. We validate the performance of the proposed model on the mobile device and the result shows it only has 49.8ms latency. The effectiveness of the proposed method on mobile devices is further supported by experimental results. A new quantitative parameter AMD (Applicability for Mobile Devices) is proposed to offer a novel approach to assist in making trade-offs for mobile devices. The proposed model obtains state-of-the-art results in terms of AMD.

Anomaly detection in surveillance videos is a challenging task due to the diversity of anomalous video content and duration. In this paper, we consider video anomaly detection as a regression problem with respect to anomaly scores of video clips under weak supervision. Hence, we propose an anomaly detection framework, called Anomaly Regression Net (ARNet), which only requires video-level labels in training stage. Further, to learn discriminative features for anomaly detection, we design a dynamic multiple-instance learning loss and a center loss for the proposed AR-Net. The former is used to enlarge the inter-class distance between anomalous and normal instances, while the latter is proposed to reduce the intra-class distance of normal instances. Comprehensive experiments are performed on a challenging benchmark: ShanghaiTech. Our method yields a new state-of-the-art result for video anomaly detection on ShanghaiTech dataset.

As surveillance technology is continuously improving, an ever-increasing number of cameras are being deployed everywhere. Relying on manual detection of anomalies through cameras may be unreliable and untimely. Therefore, the application of deep learning in video anomaly detection is being extensively studied. Anomaly Detection (AD) refers to identifying events that deviate from the desired actions. This article discusses representative unsupervised and weakly-supervised learning methods applied to various data types. In these machine learning methods, Generative Adversarial Network, Auto Encoder, Recurrent Neural Network, etc. are broadly adopted for AD. Some renowned and new datasets are reviewed. Furthermore, we also proposed several future directions of research in video anomaly detection.

Video anomaly detection plays a crucial role in various fields such as surveillance, health monitoring, and industrial quality control. This research paper introduces a novel contribution to the field by presenting MaskedConv3D layers within a modified UNet architecture. These MaskedConv3D layers provide a unique approach to information propagation in three-dimensional video data by selectively masking temporal regions of convolutional kernels. By incorporating these layers into the skip connections of the UNet, the model gains the ability to infer missing information in the temporal domain based on the surrounding context. This innovative mechanism enhances the preservation of spatial and temporal details, addressing the challenge of effectively detecting anomalies in video data. The proposed methodology is evaluated on popular video datasets, showcasing its effectiveness in capturing intricate patterns and contexts. The results highlight the superiority of the modified UNet with MaskedConv3D layers compared to traditional approaches. Overall, this research introduces a novel technique for information propagation in video data and demonstrates its potential for advancing video anomaly detection.

Video anomaly detection (VAD) is a challenging computer vision task that aims at identifying the events that do not conform to expected behaviors in a video. Training an unsupervised learning model on normal data is one of the most prevalent VAD paradigms owning to rare abnormal snippets. However, anomaly related and irrelated information sharing feature space will degrade the representation ability of the model for abnormal events. In this article, we propose a novel unsupervised video anomaly detection method, named Cross-Modal Two-Stream Target Focused Network (CMTSTFN) to focus on the anomaly related target and alleviate the impact of anomaly irrelated information on the model. Firstly, we use the global-local optical flow as a priori knowledge to blur the anomaly irrelated information in the image domain by Gaussian blur kernel. More importantly, conditioned on the blurred video frame, we then employ a dual cross-modal attention module to reconstruct the fused image feature map and optical flow feature map, respectively. Therefore, both spatial and temporal cues of abnormal events could further deteriorate the quality of the final reconstructed frame, making the anomalies more detectable. Experimental results on UCSD Pred1 and UCSD Pred2 demonstrate the effectiveness of the proposed CMTSTFN.