Normal Video Research Articles

Unsupervised Anomaly Video Detection via a Double-Flow ConvLSTM Variational Autoencoder

With the rapid increase of video surveillance points in the market in recent years, video anomaly detection has gained extensive attention in the security field. At present, the distribution of normal and anomalous data is unbalanced in unlabeled video data. Variational autoencoder (VAE), as one of the typical deep generative models, gets increasingly popular in unsupervised anomaly detection. However, this model is not good at processing time-series data, especially video data. In addition, the strong generalization ability which is over-reconstructing anomaly behavior of many autoencoder-based works leads to the missed anomaly detection. To solve these problems, in this paper, we present a double-flow convolutional long short-term memory variational autoencoder (DF-ConvLSTM-VAE) to model the probabilistic distribution of the normal video in an unsupervised learning scheme, and to reconstruct videos without anomaly objects for anomaly video detection. Experiments verify the effectiveness and competitiveness of our DF-ConvLSTM-VAE on multiple public benchmark datasets. In particular, our model achieves the state-of-the-art performance on anomalous event count.

Transferring the sandwich principle to instructional videos: is it worth the effort?

BackgroundThe sandwich principle is an educational concept that regularly alternates between collective and individual learning phases within one learning unit. Applying sandwich principle to lectures has proven to be more effective for learning outcomes than classical lectures. Supposedly, this teaching format also leads to a beneficial knowledge transfer when applied to other teaching formats. Therefore, the aim of this study was to investigate the effect of the sandwich principle on instructional videos and how its use was evaluated by students.MethodsParticipants (n = 51) were randomly allocated into two groups. Both groups were given a test to assess the baseline level of knowledge. Afterwards, the control group watched the normal instructional video on cleft lips and palates, while the sandwich group watched the same video modified according to the sandwich principle. The participants then had to answer 30 single-choice questions to assess their knowledge gain and evaluate the instructional video. Long-term retention of the knowledge was tested again 6 months later using the same test questions. The unpaired t-test and ANOVA were used to compare the results.ResultsComparison of the pre-test and post-test results of both groups showed significantly increased test scores (p < 0.0001). Regarding long-term retention, the mean test scores were still significantly higher in both groups than before watching the video (p < 0.0001). For all test results, there was no significant difference between the groups (p > 0.05). The evaluation showed that the students highly appreciated the modified video and found the interruptions for repetition of previously learned knowledge useful.ConclusionThe hypothesis that the modification of instructional videos according to the sandwich principle would lead to an improved learning outcome could not be proved subjectively or objectively. Nevertheless, the teaching format was highly appreciated by the students and may have increased their motivation to learn with instructional videos.

An abnormal event detection method based on the Riemannian manifold and LSTM network

In this paper, we propose an abnormal event detection method based on the combination of Riemannian manifold and LSTM network. We divide the video into a series of non overlapping regions, and extract HOG and HOF as original appearance and motion features. Because the existing abnormal event detection methods pay little attention to the changes rate of appearance and motion, we propose a new feature called Riemannian manifold distance feature that can represent the topological relationship of features at different moment in the same position and capture the change rate of the current feature. Once the change rate of the current features changes, the topological relationship of features will change. We use ISOMAP algorithm to embed manifold into low dimensional Euclidean space. In the training phase, the dimension-reduced features of the same position at different historical moment are used as the input of each time step of the LSTM network, and the dimension-reduced feature of the current moment is used as expected output. And in this stage, we only use the normal video, so the LSTM will learn the changing rules of the features contained in normal events. In the testing phase, we use LSTM network to predict current dimension-reduced HOG and HOF features. After that, we can obtain the Riemannian manifold distance feature of actual dimension-reduced feature and predicted dimension-reduced feature. In order to detect abnormal events, we propose an abnormal score, which is used to measure whether the appearance/ movement and the change rate of them are close to the expectation. The abnormal score can detect the abnormal event according to dimension-reduced appearance feature, dimension-reduced motion feature, appearance Riemannian manifold distance feature and motion Riemannian manifold distance feature. Experiments on four benchmark datasets show the competitive performance of our method with the state-of-the-art methods.

The Role of Virtual Reality to Enhance Cultural Communication

This paper aimed to recognize the role of virtual reality to enhance cultural communication by allowing to the participators who divided into three groups to watch the same substance in three different displays—virtual reality, normal video, images and text—to measure the process of retrieving information or events with a specific cue. The results of the study indicate that the virtual world experience enhances people's awareness of the culture. This new medium, VR, allows for participants to be immersed in the culture in ways that are not possible through the use of other media. Also, it can have an effective impact in cultural communication by allowing the participators to be virtually present at civilized events.

Semi-Supervised Anomaly Detection in Video-Surveillance Scenes in the Wild.

Surveillance cameras are being installed in many primary daily living places to maintain public safety. In this video-surveillance context, anomalies occur only for a very short time, and very occasionally. Hence, manual monitoring of such anomalies may be exhaustive and monotonous, resulting in a decrease in reliability and speed in emergency situations due to monitor tiredness. Within this framework, the importance of automatic detection of anomalies is clear, and, therefore, an important amount of research works have been made lately in this topic. According to these earlier studies, supervised approaches perform better than unsupervised ones. However, supervised approaches demand manual annotation, making dependent the system reliability of the different situations used in the training (something difficult to set in anomaly context). In this work, it is proposed an approach for anomaly detection in video-surveillance scenes based on a weakly supervised learning algorithm. Spatio-temporal features are extracted from each surveillance video using a temporal convolutional 3D neural network (T-C3D). Then, a novel ranking loss function increases the distance between the classification scores of anomalous and normal videos, reducing the number of false negatives. The proposal has been evaluated and compared against state-of-art approaches, obtaining competitive performance without fine-tuning, which also validates its generalization capability. In this paper, the proposal design and reliability is presented and analyzed, as well as the aforementioned quantitative and qualitative evaluation in-the-wild scenarios, demonstrating its high sensitivity in anomaly detection in all of them.

PornNet: A Unified Deep Architecture for Pornographic Video Recognition

In the era of big data, massive harmful multimedia resources publicly available on the Internet greatly threaten children and adolescents. In particular, recognizing pornographic videos is of great importance for protecting the mental and physical health of the underage. In contrast to the conventional methods which are only built on image classifier without considering audio clues in the video, we propose a unified deep architecture termed PornNet integrating dual sub-networks for pornographic video recognition. More specifically, with image frames and audio clues extracted from the pornographic videos from scratch, they are respectively delivered to two deep networks for pattern discrimination. For discriminating pornographic frames, we propose a local-context aware network that takes into account the image context in capturing the key contents, whilst leveraging an attention network which can capture temporal information for recognizing pornographic audios. Thus, we incorporate the recognition scores generated from the two sub-networks into a unified deep architecture, while making use of a pre-defined aggregation function to produce the whole video recognition result. The experiments on our newly-collected large dataset demonstrate that our proposed method exhibits a promising performance, achieving an accuracy at 93.4% on the dataset including 1 k pornographic samples along with 1 k normal videos and 1 k sexy videos.

Abnormal event detection algorithm based on dual attention future frame prediction and gap fusion discrimination

Rapid and accurate detection of crowd abnormal events, such as stampedes and violent attacks in public places, has great research significance and application value. Due to the diversity and uncertainty of abnormal events, almost all existing methods tackle the problem by minimizing the reconstruction errors of training data, which cannot guarantee a larger reconstruction error for all abnormal events. According to the idea that “normal events can be predicted, abnormal events cannot be predicted,” we proposed a future frame prediction-based anomaly detection algorithm. First, the generative adversarial network (GAN) is trained by the normal videos to predict normal future frames. Then, it can determine the existence of abnormal events by identifying the difference between the ground truth and predicted video frame. In the design of the GAN, the attention module is introduced to improve the prediction level of the network. At the same time, the optical flow information is added for motion constraint to improve the constraint ability on the appearance characteristics. In the testing stage, the appearance gap and optical flow gap between the ground truth and the predicted video frame are fused to determine whether the frame is abnormal. The experimental results on the datasets of CUHK Avenue, UCSD, and ShanghaiTech show that the proposed algorithm is superior to that of the current mainstream anomaly detection algorithms.

Gaussian-Wiener Representation and Hierarchical Coding Scheme for Focal Stack Images

Focal stack images (FoSIs) are a set of 2D images that captured one scene with serial focal depths. The redundancies of FoSIs mainly come from gradual focused depths changes rather than motion of objects. Conventional coding schemes cannot fully exploit such redundancies, leading to coding inefficiency. In this paper, we propose a new Gaussian-Wiener representation to model the gradual focused depths changes among FoSIs. In the representation, image degradation-restoration relations are utilized to describe the focus-defocus changing characteristics of FoSIs. Based on this representation, we propose a new hierarchical coding scheme for fully exploiting the inter-frame redundancies of FoSIs. In the scheme, a Gaussian-Wiener representation based inter prediction (GWR-IP) is presented by embedding Gaussian convolution and Wiener deconvolution into normal video encoder. Block-wise focus-defocus changing of FoSIs can be predicted in bi-directional manner by solving optimization problem. For higher coding efficiency, a Gaussian-Wiener representation based hierarchical prediction structure (GWR-HPS) is also designed and applied in the coding scheme. The proposed coding scheme is performed on 10 test sequences, including 5 synthetic scenes and 5 realistic scenes. Experimental results show that proposed coding scheme can obtain 2.640 dB PSNR gains and 51.830% bitrate savings on average of all test sequences in Low Delay P configuration, 2.123 dB PSNR gains and 43.975% bitrate savings in Low Delay B configuration, and 1.044 dB PSNR gains and 26.078% bitrate savings in Random Access configuration. Particularly, it achieves up to 65.544% bit rate savings and 3.901 dB PSNR increments for test sequence I09 in Low Delay P configuration. Furthermore, ablation test demonstrates that Gaussian representation contributes more on coding performance than Wiener representation and GWR-HPS.

Real time face recognition of video surveillance system using haar cascade classifier

&lt;div&gt;This project investigates the use of face recognition for a surveillance system. The normal video surveillance system uses in closed-circuit television (CCTV) to record video for security purpose. It is used to identify the identity of a person through their appearances on the recorded video, manually. Today’s video surveillance camera system usually not occupied with a face recognition system. With some modification, a surveillance camera system can be used as face detection and recognition that can be done in real-time. The proposed system makes use of surveillance camera system that can identify the identity of a person automatically by using face recognition of Haar cascade classifier. The hardware used for this project were Raspberry Pi as a processor and Pi Camera as a camera module. The development of this project consist of three main phases which were data gathering, training recognizer, and face recognition process. All three phases have been executed using Python programming and OpenCV library, which have been performed in a Raspbian operation system. From the result, the proposed system successfully displays the output result of human face recognition, with facial angle within ±40°, in medium and normal light condition, and within a distance of 0.4 to 1.2 meter. Targeted image are allowed to wear face accessory as long as not covering the face structure. In conclusion, this system considered, can reduce the cost of manpower in order to identify the identity of a person in real time situation.&lt;/div&gt;

A Skill-Based Visual Attention Model for Cloud Gaming

Despite its recent advances and increasing industrial interest, cloud gaming’s high bandwidth usage is still one of its major challenges. In this paper, we demonstrate how incorporating visual attention into cloud gaming helps to reduce bitrate without negatively affecting the player’s quality of experience. We show that current visual attention models, which work well for normal videos, underperform in the context of cloud gaming videos. Hence, we propose our novel model, by developing a skill-based visual attention model, based on a cloud gaming dataset. First, it is demonstrated how players’ attention maps are correlated with their skill levels and how this can be exploited to improve the accuracy of visual attention modeling. Then, this fact is used to cluster attention maps, according to the player’s skill level. A simple yet effective method is introduced to predict players’ skill levels using their performance in game. Finally, the models are incorporated into the video encoder to perceptually optimize the bitrate allocation. Incorporating the player’s skill level into our model improves the accuracy of saliency maps by 14% with respect to the baseline, and 24% with respect to competing methods, in terms of Normalized Scanpath Saliency (NSS). Furthermore, we show that the maximum possible amount of video bitrate reduction depends on the player’s skill level. Experimental results show 13%, 5%, and 15% reduction in video bitrate for beginner, intermediate, and expert players, respectively.

Oculomotor Responses to Dynamic Stimuli in a 44-Channel Suprachoroidal Retinal Prosthesis.

PurposeTo investigate oculomotor behavior in response to dynamic stimuli in retinal implant recipients.MethodsThree suprachoroidal retinal implant recipients performed a four-alternative forced-choice motion discrimination task over six sessions longitudinally. Stimuli were a single white bar (“moving bar”) or a series of white bars (“moving grating”) sweeping left, right, up, or down across a 42″ monitor. Performance was compared with normal video processing and scrambled video processing (randomized image-to-electrode mapping to disrupt spatiotemporal structure). Eye and head movement was monitored throughout the task.ResultsTwo subjects had diminished performance with scrambling, suggesting retinotopic discrimination was used in the normal condition and made smooth pursuit eye movements congruent to the moving bar stimulus direction. These two subjects also made stimulus-related eye movements resembling optokinetic reflex (OKR) for moving grating stimuli, but the movement was incongruent with stimulus direction. The third subject was less adept at the task, appeared primarily reliant on head position cues (head movements were congruent to stimulus direction), and did not exhibit retinotopic discrimination and associated eye movements.ConclusionsOur observation of smooth pursuit indicates residual functionality of cortical direction-selective circuits and implies a more naturalistic perception of motion than expected. A distorted OKR implies improper functionality of retinal direction-selective circuits, possibly due to retinal remodeling or the non-selective nature of the electrical stimulation.Translational RelevanceRetinal implant users can make naturalistic eye movements in response to moving stimuli, highlighting the potential for eye tracker feedback to improve perceptual localization and image stabilization in camera-based visual prostheses.

A new method of abnormal behavior detection using LSTM network with temporal attention mechanism

In this paper, we propose an accurate and effective method for detecting abnormal behavior. We consider the video as a series of frame sequences; in the training phase, our deep learning framework is used to extract appearance features and learn the relationship between historical features and current features in the normal video. In the testing phase, the predicted features that differ from the actual features are considered as abnormal. Our model is designed as a feature prediction framework with a new temporal attention mechanism. In the feature extraction stage, we transform a pre-trained Vgg16 network into a fully convolutional neural network and used the third pooling layer output as the appearance feature extraction to effectively capture static appearance features. Then, a new temporal attention mechanism is introduced to learn the contribution of different historical appearance features at the same position to the current features, thereby solving the problem of representing dynamic motion features. Finally, the LSTM network is used to decode the historical feature sequences with temporal attention to predict the features at the current moment. Those actual features that differ from the predicted features are considered as abnormal features. Using upsampling for the abnormal features locates abnormal behavior on the original frames. Experiments on two benchmark datasets demonstrate the competitive performance of our method with the state-of-the-art methods.

Fast sparse coding networks for anomaly detection in videos

The semi-supervised video anomaly detection assumes that only normal video clips are available for training. Therefore, the intuitive idea is either to learn a dictionary by sparse coding or to train encoding-decoding neural networks by minimizing the reconstruction errors. For the former, the optimization of sparse coefficients is extremely time-consuming. For the latter, this manner cannot guarantee that an abnormal data corresponds to a larger reconstruction error due to the strong generalization of neural networks. To remedy their weaknesses and leverage their strengths, we propose a Fast Sparse Coding Network (FSCN) based on High-level Features. First, we propose a two-stream neural network to extract Spatial-Temporal Fusion Features (STFF) in hidden layers. With the STFF at hand, we use a Fast Sparse Coding Network to build a normal dictionary. By leveraging the predictor to produce approximate sparse coefficients, our FSCN generates sparse coefficients within a forward pass, which is simple and computationally efficient. Compared with traditional sparse coding based methods, FSCN is hundreds of or even thousands of times faster at the test stage. Extensive experiments on benchmark datasets demonstrate that our method reaches the state-of-the-art level.11Code will be released at https://github.com/Roc-Ng/FSCN_AnomalyDetection.

Real Life Human Movement Realization in Multimodal Group Communication Using Depth Map Information and Machine Learning

The latest advancements in the evolution of depth map information’s has paved way for interesting works like object recognition sign detection and human movement detection etc. The real life human movement detection or their activity identification is very challenging and tiresome. Since the real life activities of the humans could be of much interest in almost all areas, the subject of identifying the human activities has gained significance and has become a most popular research field. Identifying the human movements /activities in the public places like airport, railways stations, hospital, home for aged become very essential due to the several benefits incurred form the human movement recognition system such as surveillance camera, monitoring devices etc. since the changes in the space and the time parameters can provide an effective way of presenting the movements, yet in the case of natural color vision, as the flatness is depicted in almost all portions of images. So the work laid out in the paper in order to identify the human movement in the real life employs the space and the time depth particulars (Spatial-Temporal depth details –STDD) and the random forest in the final stage for movement classification. The technology put forth utilize the Kinect sensors to collecting the information’s in the data gathering stage. The mechanism laid out to identify the human movements is test with the MATLAB using the Berkley and the Cornell datasets. The mechanism proposed through the acquired results proves to deliver a better performance compared to the human movements captured using the normal video frames.

A working human abnormal operation recognition approach based on the deep multi-instance sorting model

Modeling abnormal behavior detection in the intelligent video monitoring system for recognition or detection of special event has attracted significant research interest in recent years. In order to achieve more effective recognition of abnormal behavior detection of the intelligent video surveillance system, this paper proposes a working human abnormal operation recognition approach based on deep multi-instance sorting model. Firstly, the uncut long video is sparse to obtain normal and abnormal behavior video segments, and the RGB and optical flow features in the segment are extracted by the deep convolution network. Video feature vectors are obtained by the consensus function and feature extractor. Then, multi-instance sorting learning is used to assign abnormality scores between 0 and 1 to feature vectors. When the abnormal values of abnormal packets are higher than those of normal packets, the abnormality score is returned and the high abnormality score is determined as an abnormal behavior. The experiments on the open THUMOS14 data set and our own XAGCWD data set using CUDA GPU accelerated computing to demonstrate that our approach improves the recognition accuracy about 10.8% and high accuracy of abnormal detection. The main purpose of this research is to apply the model proposed in this paper to the intelligent workshop behavior monitoring system to effectively realize the safety management of workshop personnel.

Real-Time Application for Generating Multiple Experiences from 360° Panoramic Video by Tracking Arbitrary Objects and Viewer’s Orientations

We propose a novel authoring and viewing system for generating multiple experiences with a single 360° video and efficiently transferring these experiences to the user. An immersive video contains much more interesting information within the 360° environment than normal videos. There can be multiple interesting areas within a 360° frame at the same time. Due to the narrow field of view in virtual reality head-mounted displays, a user can only view a limited area of a 360° video. Hence, our system is aimed at generating multiple experiences based on interesting information in different regions of a 360° video and efficient transferring of these experiences to prospective users. The proposed system generates experiences by using two approaches: (1) Recording of the user’s experience when the user watches a panoramic video using a virtual reality head-mounted display, and (2) tracking of an arbitrary interesting object in a 360° video selected by the user. For tracking of an arbitrary interesting object, we have developed a pipeline around an existing simple object tracker to adapt it for 360° videos. This tracking algorithm was performed in real time on a CPU with high precision. Moreover, to the best of our knowledge, there is no such existing system that can generate a variety of different experiences from a single 360° video and enable the viewer to watch one 360° visual content from various interesting perspectives in immersive virtual reality. Furthermore, we have provided an adaptive focus assistance technique for efficient transferring of the generated experiences to other users in virtual reality. In this study, technical evaluation of the system along with a detailed user study has been performed to assess the system’s application. Findings from evaluation of the system showed that a single 360° multimedia content has the capability of generating multiple experiences and transfers among users. Moreover, sharing of the 360° experiences enabled viewers to watch multiple interesting contents with less effort.

Anomaly Detection Based on Locality Sensitive Hashing with Genetic Algorithm

With the increase of public safety awareness, video anomaly detection has attracted researchers’ attention. In the paper, a novel approach is proposed to detect anomalies in the video. It is based on Locality Sensitive Hashing (LSH), which maps similar data to the same bucket with high probabilities, and non-similar data is mapped to the same bucket with a low probability to detect abnormal videos that are not similar to normal videos. In order to improve the probability of similar data mapping into the same bucket, the Genetic Algorithm (GA) is used to optimize the entire hash function group while maintaining the diversity of the hash function group. The algorithm gets AUC 0.78 on the dataset UCSD ped1 and AUC 0.94 on the dataset UCSD ped2, which confirmed the effectiveness of the algorithm.

CAD-CAP: a 25,000-image database serving the development of artificial intelligence for capsule endoscopy.

Background and study aims Capsule endoscopy (CE) is the preferred method for small bowel (SB) exploration. With a mean number of 50,000 SB frames per video, SBCE reading is time-consuming and tedious (30 to 60 minutes per video). We describe a large, multicenter database named CAD-CAP (Computer-Assisted Diagnosis for CAPsule Endoscopy, CAD-CAP). This database aims to serve the development of CAD tools for CE reading. Materials and methods Twelve French endoscopy centers were involved. All available third-generation SB-CE videos (Pillcam, Medtronic) were retrospectively selected from these centers and deidentified. Any pathological frame was extracted and included in the database. Manual segmentation of findings within these frames was performed by two pre-med students trained and supervised by an expert reader. All frames were then classified by type and clinical relevance by a panel of three expert readers. An automated extraction process was also developed to create a dataset of normal, proofread, control images from normal, complete, SB-CE videos. Results Four-thousand-one-hundred-and-seventy-four SB-CE were included. Of them, 1,480 videos (35 %) containing at least one pathological finding were selected. Findings from 5,184 frames (with their short video sequences) were extracted and delimited: 718 frames with fresh blood, 3,097 frames with vascular lesions, and 1,369 frames with inflammatory and ulcerative lesions. Twenty-thousand normal frames were extracted from 206 SB-CE normal videos. CAD-CAP has already been used for development of automated tools for angiectasia detection and also for two international challenges on medical computerized analysis.

Dual Discriminator Generative Adversarial Network for Video Anomaly Detection

Video anomaly detection is an essential task because of its numerous applications in various areas. Because of the rarity of abnormal events and the complicated characteristic of videos, video anomaly detection is challenging and has been studied for a long time. In this paper, we propose a semi-supervised approach with a dual discriminator-based generative adversarial network structure. Our method considers more motion information in video clips compared with previous approaches. Specifically, in the training phase, we predict future frames for normal events via a generator and attempt to force the predicted frames to be similar to their ground truths. In addition, we utilize both a frame discriminator and motion discriminator to adverse the generator to generate more realistic and consecutive frames. The frame discriminator attempts to determine whether the input frames are generated or original frames sampled from the normal video. The motion discriminator attempts to determine whether the given optical flows are real or fake. Fake optical flows are estimated from generated frames and adjacent frames, and real optical flows are estimated from the real frames sampled from original videos. Then, in the testing phase, we evaluate the quality of predicted frames to obtain the regular score, and we consider those frames with lower prediction qualities as abnormal frames. The results of experiments on three publicly available datasets demonstrate the effectiveness of our proposed method.

P203 Unsupervised one-class classification and anomaly detection of stress echocardiograms with deep denoising spatio-temporal autoencoders

Abstract Introduction The combination of medical knowledge, experience and AI algorithms have supported the advancement of patient care and the lowering of healthcare costs. Machine and deep learning methods enable the extraction of meaningful patterns that remain beyond human perception. Numerous computer-aided diagnosis and detection systems have been developed to assist in the assessment of stress echocardiograms. However, issues are encountered when facing imbalanced, limited, and unannotated datasets. Learning from imbalanced medical datasets impairs diagnostic accuracy due to classifier bias and overfitting. Furthermore, datasets comprising of all existing abnormal classes are impossible to obtain, hence supervised algorithms would fail to generate predictions for classes devoid of training samples. Moreover, reliance on prior knowledge in the form of expert annotation and anatomical region extraction impairs scalability, as these procedures are time-consuming, computationally expensive, and limited to specific tasks. Purpose We aimed to perform one-class classification and anomaly detection of stress echocardiograms using unsupervised deep learning techniques to discriminate between normal and abnormal videos as well as to localise wall motion abnormalities within individual frames. Methods Deep denoising spatio-temporal autoencoder networks were employed to learn visual and motion representations from multiple echocardiographic cross-sections and stress stages. Extracted middle layer features were modelled by one-class support vector machines to discriminate between regular and irregular echocardiograms despite the absence of abnormal training samples. Reconstruction errors allowed for direct visualisation and localisation of anomalous cardiac regions, without the need for annotated training data or segmentation of structures. Results 2D B-mode stress echocardiograms acquired from 36 patients were classified as normal or abnormal based on patient reports and served as the ground truth. Results revealed that learnt features extracted from spatio-temporal autoencoders trained solely on normal data can be utilised to classify abnormal echocardiograms with a high level of accuracy, sensitivity and specificity. In addition to that, as validated by an expert reader, spatio-temporal autoencoder reconstruction errors were capable of detecting and localising wall motion abnormalities in specific cardiac regions without prior knowledge of abnormal segments. Conclusions The trained model enables the classification and detection of spatio-temporal abnormalities in stress echocardiograms. Therefore, the proposed networks have the potential of assisting in the global and regional assessment of stress echocardiograms.