Sequence Of Frames Research Articles

Major histocompatibility complex class IIα and IIβ of pufferfish (Takifugu obscurus): Identification and functional characterization in response to bacterial challenge.

Classical major histocompatibility complex (MHC) class II molecules play an essential role in immune system. In this study, MHC IIα (Pf-MHC IIα) and MHC IIβ (Pf-MHC IIβ) homology genes from pufferfish (Takifugu obscurus) were cloned and their functional characterization in response to bacterial challenge was identified. The nucleotide sequences of the open reading frames (ORFs) of pufferfish Pf-MHC IIα and Pf-MHC IIβ were 708 bp and 750 bp, encoding 235 aa and 249 aa, respectively. The structure of Pf-MHC IIα or Pf-MHC IIβ contained a signal peptide, an α1/β1 domain, an α2/β2 domain, a transmembrane region and a cytoplasmic region. Multiple sequence alignment and phylogenetic analysis showed that Pf-MHC IIα and Pf-MHC IIβ molecules had the highest similarity with Fugu rubripes (Takifugu rubripes). Cellular localization analysis indicated that the distribution of Pf-MHC IIα and Pf-MHC IIβ was in the lymphocyte membrane and cytoplasm. qRT-PCR results showed that Pf-MHC IIα and Pf-MHC IIβ expressed relatively high in skin, gills and gut. In addition, after stimulation challenge in vitro (lipopolysaccharide, or polyinosinic: polycytidylic acid) and in vivo (A. hydrophila), the mRNA expressions of Pf-MHC IIα and Pf-MHC IIβ were significantly up-regulated in lymphocytes and in tissues of skin, gills, gut and head kidney. Moreover, Pf-MHC IIα or Pf-MHC IIβ neutralization reduced the ability of A. hydrophila to induce the expressions of lymphocyte cytokines (TNF-α, IL-1β and IL-10). Overall, it is speculated that Pf-MHC IIα and Pf-MHC IIβ may play an important role in the host response against A. hydrophila in pufferfish.

The concept and essence of a cinematographic work as an object of legal protection in copyright law

In this article, the author reveals the legal nature and essence of a cinematographic work as an object of copyright. It is noted that the current legislation of Ukraine does not have a legislative definition of a cinematographic work, but the Law of Ukraine "On Copyright and Related Rights" proposes a definition of an audiovisual work, which cannot be equated with a cinematographic work. It should be noted that many important international legal documents in the field of copyright protection operate with the concept of cinematographic works and give its definition (Bern Convention for the Protection of Literary and Artistic Works, European Convention on Joint Cinematographic Production, etc.). A number of foreign countries also have a statutory definition of cinematographic works (Canada, France, etc.). On the basis of definitions of cinematographic works existing in the doctrine of intellectual property law, their copyright characteristics and the current legislation of Ukraine, it is proposed to distinguish general and special features of cinematographic works. The former are manifested in all works, including cinematographic works, and the latter objectively reveal the essence of cinematographic works. In particular, the author proposes to include the following: 1) the cinematographic work is presented in the form of frames; 2) frames are consecutive; 3) the frames are combined with a creative idea; 4) the work is available for perception using technical means; 5) the sequence of frames forms a moving image. The relationship between these features is illustrated, where some follow from others. The structure of the cinematographic work is analyzed, where its protected and unprotected elements are distinguished. It is concluded that a cinematographic work can be classified as a complex work among the objects of copyright, and therefore, for the proper exploitation of a cinematographic work, it is necessary that its producers observe non-property and property rights to those objects of intellectual property law that were used during the creation movie.

A spatio-temporal network for video semantic segmentation in surgical videos.

Semantic segmentation in surgical videos has applications in intra-operative guidance, post-operative analytics and surgical education. Models need to provide accurate predictions since temporally inconsistent identification of anatomy can hinder patient safety. We propose a novel architecture for modelling temporal relationships in videos to address these issues. We developed a temporal segmentation model that includes a static encoder and a spatio-temporal decoder. The encoder processes individual frames whilst the decoder learns spatio-temporal relationships from frame sequences. The decoder can be used with any suitable encoder to improve temporal consistency. Model performance was evaluated on the CholecSeg8k dataset and a private dataset of robotic Partial Nephrectomy procedures. Mean Intersection over Union improved by 1.30% and 4.27% respectively for each dataset when the temporal decoder was applied. Our model also displayed improvements in temporal consistency up to 7.23%. This work demonstrates an advance in video segmentation of surgical scenes with potential applications in surgery with a view to improve patient outcomes. The proposed decoder can extend state-of-the-art static models, and it is shown that it can improve per-frame segmentation output and video temporal consistency.

Alternative Reading Frames are an Underappreciated Source of Protein Sequence Novelty.

Protein-coding DNA sequences can be translated into completely different amino acid sequences if the nucleotide triplets used are shifted by a non-triplet amount on the same DNA strand or by translating codons from the opposite strand. Such "alternative reading frames" of protein-coding genes are a major contributor to the evolution of novel protein products. Recent studies demonstrating this include examples across the three domains of cellular life and in viruses. These sequences increase the number of trials potentially available for the evolutionary invention of new genes and also have unusual properties which may facilitate gene origin. There is evidence that the structure of the standard genetic code contributes to the features and gene-likeness of some alternative frame sequences. These findings have important implications across diverse areas of molecular biology, including for genome annotation, structural biology, and evolutionary genomics.

Motion vector steganography algorithm of sports training video integrating with artificial bee colony algorithm and human-centered AI for web applications

AbstractIn multimedia correspondence, steganography schemes are commonly applied. To reduce storage capacity, multimedia files, including images, are always compressed. Most steganographic video schemes are, therefore, not compression tolerant. In the frame sequences, the video includes extra hidden space. Artificial intelligence (AI) creates a digital world of real-time information for athletes, sponsors, and broadcasters. AI is reshaping business, and although it has already produced a significant impact on other sectors, the sports industry is the newest and most receptive one. Human-centered AI for web applications has substantially influenced audience participation, strategic plan execution, and other aspects of the sports industry that have traditionally relied heavily on statistics. Thus, this study presents the motion vector steganography of sports training video integrating with the artificial bee colony algorithm (MVS-ABC). The motion vector stenography detects the hidden information from the motion vectors in the sports training video bitstreams. Artificial bee colony (ABC) algorithm optimizes the block assignment to inject a hidden message into a host video, in which the block assignment is considered a combinatorial optimization problem. The experimental analysis evaluates the data embedding performance using steganographic technology compared with existing embedding technologies, using the ABC algorithm compared with other genetic algorithms. The findings show that the proposed model can give the highest performance in terms of embedding capacity and the least error rate of video steganography compared with the existing models.

Fusion of Multi-Modal Features to Enhance Dense Video Caption

Dense video caption is a task that aims to help computers analyze the content of a video by generating abstract captions for a sequence of video frames. However, most of the existing methods only use visual features in the video and ignore the audio features that are also essential for understanding the video. In this paper, we propose a fusion model that combines the Transformer framework to integrate both visual and audio features in the video for captioning. We use multi-head attention to deal with the variations in sequence lengths between the models involved in our approach. We also introduce a Common Pool to store the generated features and align them with the time steps, thus filtering the information and eliminating redundancy based on the confidence scores. Moreover, we use LSTM as a decoder to generate the description sentences, which reduces the memory size of the entire network. Experiments show that our method is competitive on the ActivityNet Captions dataset.

Human Action Recognition Using Key-Frame Attention-Based LSTM Networks

Human action recognition is a classical problem in computer vision and machine learning, and the task of effectively and efficiently recognising human actions is a concern for researchers. In this paper, we propose a key-frame-based approach to human action recognition. First, we designed a key-frame attention-based LSTM network (KF-LSTM) using the attention mechanism, which can be combined with LSTM to effectively recognise human action sequences by assigning different weight scale values to give more attention to key frames. In addition, we designed a new key-frame extraction method by combining an automatic segmentation model based on the autoregressive moving average (ARMA) algorithm and the K-means clustering algorithm. This method effectively avoids the possibility of inter-frame confusion in the temporal sequence of key frames of different actions and ensures that the subsequent human action recognition task proceeds smoothly. The dataset used in the experiments was acquired with an IMU sensor-based motion capture device, and we separately extracted the motion features of each joint using a manual method and then performed collective inference.

Optimized Small Waterbird Detection Method Using Surveillance Videos Based on YOLOv7

Waterbird monitoring is the foundation of conservation and management strategies in almost all types of wetland ecosystems. China’s improved wetland protection infrastructure, which includes remote devices for the collection of larger quantities of acoustic and visual data on wildlife species, increased the need for data filtration and analysis techniques. Object detection based on deep learning has emerged as a basic solution for big data analysis that has been tested in several application fields. However, these deep learning techniques have not yet been tested for small waterbird detection from real-time surveillance videos, which can address the challenge of waterbird monitoring in real time. We propose an improved detection method by adding an extra prediction head, SimAM attention module, and sequential frame to YOLOv7, termed as YOLOv7-waterbird, for real-time video surveillance devices to identify attention regions and perform waterbird monitoring tasks. With the Waterbird Dataset, the mean average precision (mAP) value of YOLOv7-waterbird was 67.3%, which was approximately 5% higher than that of the baseline model. Furthermore, the improved method achieved a recall of 87.9% (precision = 85%) and 79.1% for small waterbirds (defined as pixels less than 40 × 40), suggesting a better performance for small object detection than the original method. This algorithm could be used by the administration of protected areas or other groups to monitor waterbirds with higher accuracy using existing surveillance cameras and can aid in wildlife conservation to some extent.

Emotion recognition at a distance: The robustness of machine learning based on hand-crafted facial features vs deep learning models

Emotion estimation from face expression analysis is nowadays a widely-explored computer vision task. In turn, the classification of expressions relies on relevant facial features and their dynamics. Despite the promising accuracy results achieved in controlled and favorable conditions, the processing of faces acquired at a distance, entailing low-quality images, still suffers from a significant performance decrease. In particular, most approaches and related computational models become extremely unstable in the case of the very small amount of useful pixels that is typical in these conditions. Therefore, their behavior should be investigated more carefully. On the other hand, real-time emotion recognition at a distance may play a critical role in smart video surveillance, especially when controlling particular kinds of events, e.g., political meetings, to try to prevent adverse actions.This work compares facial expression recognition at a distance by: 1) a deep learning architecture based on state-of-the-art (SOTA) proposals, which exploits the whole images to autonomously learn the relevant embeddings; 2) a machine learning approach that relies on hand-crafted features, namely the facial landmarks preliminarily extracted using the popular Mediapipe framework. Instead of using either the complete sequence of frames or only the final still image of the expression, like current SOTA approaches, the two proposed methods are designed to use rich temporal information to identify three different stages of emotion. Expressions are time-split accordingly into four phases to better exploit their temporal-dependent dynamics. Experiments were conducted on the popular Extended Cohn-Kanade dataset (CK+). It was chosen for its wide use in related literature, and because it includes videos of facial expressions and not only still images. The results show that the approach relying on machine learning via hand-crafted features is more suitable for classifying the initial phases of the expression and does not decay in terms of accuracy when images are at a distance (only 0.08% of decay). On the contrary, deep learning not only has difficulties classifying the initial phases of the expressions but also suffers from relevant performance decay when considering images at a distance (52.68% accuracy decay).

TWISTED SHIFT-INVARIANT SYSTEM IN

AbstractWe consider a general twisted shift-invariant system, $V^{t}(\mathcal {A})$ , consisting of twisted translates of countably many generators and study the problem of obtaining a characterization for the system $V^{t}(\mathcal {A})$ to form a frame sequence or a Riesz sequence. We illustrate our theory with some examples. In addition to these results, we study a dual twisted shift-invariant system and also obtain an orthonormal sequence of twisted translates from a given Riesz sequence of twisted translates.

Reinforcement Learning for Pass Detection and Generation of Possession Statistics in Soccer

We propose a reinforcement learning (RL) based technique to detect passes from the video of a soccer match. The detection of passes determines ball possession statistics of a soccer match. A sequence of video frames is mapped to a sequence of states, such as ball with team A or team B or ball not possessed either by team A or B. The agent of RL learns the frame-to-state mapping and the optimal policy to decide the mapping task. We propose a novel reward function by utilizing contextual information of the soccer game in order to help the agent decide the optimal policy. In this context, the advantage of RL is in the integration of a reward system in choosing an action that maps a video frame of a soccer match to one of three possible states. Unlike competing methods, we design the RL model in a way so that explicit identification of team labels of players is not required. We introduce a Deep Recurrent Q-Network (DRQN) to learn the optimal policy. For efficient training of the DRQN, we have proposed de-correlated experience replay (DER), a strategy that selects important experiences based on the correlations of the experiences stored in the replay memory. Experimental results show that at least 5.75% and 2.1% better accuracy are achieved in calculating pass detection and possession statistics, respectively, compared to similar approaches.

A novel GPU based Geo-Location Inference Attack on WebGL framework

In the past few years, graphics processing units (GPUs) have become an indispensable part of modern computer systems, not only for graphics rendering but also for intensive parallel computing. Given that many tasks running on GPUs contain sensitive information, security concerns have been raised, especially about potential GPU information leakage. Previous works have shown such concerns by showing that attackers can use GPU memory allocations or performance counters to measure victim side effects. However, such an attack has a critical drawback that it requires a victim to install desktop applications or mobile apps yielding it uneasy to be deployed in the real world. In this paper, we solve this drawback by proposing a novel GPU-based side-channel Geo-Privacy inference attack on the WebGL framework, namely, GLINT (stands for Geo-Location Inference Attack). GLINT merely utilizes a lightweight browser extension to measure the time elapsed to render a sequence of frames on well-known map websites, e.g., Google Maps, or Baidu Maps. The measured stream of time series is then employed to infer geologically privacy-sensitive information, such as a search on a specific location. Upon retrieving the stream, we propose a novel online segmentation algorithm for streaming data to determine the start and end points of privacy-sensitive time series. We then combine the DTW algorithm and KNN algorithm on these series to conclude the final inference on a user’s geo-location privacy.We conducted real-world experiments to testify our attack. The experiments show that GeoInfer can correctly infer more than 83% of user searches regardless of the locations and map websites, meaning that our Geo-Privacy inference attack is accurate, practical, and robust. To counter this attack, we implemented a defense strategy based on Differential Privacy to hinder obtaining accurate rendering data. We found that this defense mechanism managed to reduce the average accuracy of the attack model by more than 70%, indicating that the attack was no longer effective. We have fully implemented GLINT and open-sourced it for future follow-up research.

Differential sense and antisense expression profiles of Syrista parreyssi (Hymenoptera: Cephidae) mitochondrial transcripts.

The transcription of the mitogenome shows a unique pattern that is both similar to and different from the nuclear and bacterial patterns. Mitochondrial transcription generates five polycistronic units from three promoters in Drosophila melanogaster, and different expression levels of genes were observed in both different and, interestingly, the same polycistronic units in D. melanogaster. This study was conducted to test this phenomenon in the mitogenome of Syrista parreyssi (Hymenoptera: Cephidae). RNA isolation and DNase digestion were performed using only one whole individual, and real-time polymerase chain reaction analyses were performed with complementary DNAs of 11 gene regions using gene-specific primers. It was found that the expression level of each gene exhibited differences from each other, and some genes (e.g., cox genes, and rrnS) were interestingly expressed at significant levels in the corresponding antisense chain. Additionally, the mitogenome of S. parreyssi was found to have the capacity to encode 169 additional peptides from 13 known protein-coding genes, most of which were encoded in antisense transcript units. One of the unique findings was a potential open reading frame sequence that was potentially encoded in the antisense rrnL gene and included a conserved cox3 domain.

Variation within major internal repeats ofKSHV invivo.

Kaposi's sarcoma-associated herpesvirus (KSHV) is the etiologic agent of Kaposi's sarcoma (KS), yet the viral genetic factors that lead to the development of KS in KSHV-infected individuals have not been fully elucidated. Nearly, all previous analyses of KSHV genomic evolution and diversity have excluded the three major internal repeat regions: the two origins of lytic replication, internal repeats 1 and 2 (IR1 and IR2), and the latency-associated nuclear antigen (LANA) repeat domain (LANAr). These regions encode protein domains that are essential to the KSHV infection cycle but have been rarely sequenced due to their extended repetitive nature and high guanine and cytosine (GC) content. The limited data available suggest that their sequences and repeat lengths are more heterogeneous across individuals than in the remainder of the KSHV genome. To assess their diversity, the full-length IR1, IR2, and LANAr sequences, tagged with unique molecular identifiers (UMIs), were obtained by Pacific Biosciences' single-molecule real-time sequencing (SMRT-UMI) from twenty-four tumors and six matching oral swabs from sixteen adults in Uganda with advanced KS. Intra-host single-nucleotide variation involved an average of 0.16 per cent of base positions in the repeat regions compared to a nearly identical average of 0.17 per cent of base positions in the remainder of the genome. Tandem repeat unit (TRU) counts varied by only one from the intra-host consensus in a majority of individuals. Including the TRU indels, the average intra-host pairwise identity was 98.3 per cent for IR1, 99.6 per cent for IR2 and 98.9 per cent for LANAr. More individuals had mismatches and variable TRU counts in IR1 (twelve/sixteen) than in IR2 (two/sixteen). There were no open reading frames in the Kaposin coding sequence inside IR2 in at least fifty-five of ninety-six sequences. In summary, the KSHV major internal repeats, like the rest of the genome in individuals with KS, have low diversity. IR1 was the most variable among the repeats, and no intact Kaposin reading frames were present in IR2 of the majority of genomes sampled.

A survey of Face detection and Recognition system

In recent days, our world has entered the age of advanced digital systems, which means humans has to interact with technology in all scopes of live, this leads too as a result to another main concept, security. There are many security technologies that was developed to keep user information save. for example, there are several technologies used to let just the accurate user to log into devices like fingerprint, face recognition, or any another type of these systems. These technologies are in continuous development and are still suffering from accuracy, so several studies were performed to get the best accuracy for any of these systems. Computer vision was developed to give the devices the ability to see and recognize objects in an image or a video scene. It is a science that aims to build an intelligent application that are able to understand the content inside images as humans can do. To create a system like this, first step is to make Data acquisition, this task is done by cameras which gives us a sequence of frames when we record a video. After that, this data is analyzed and studied to extract important features from it. This feature will allow us to re- construct the description of the outer world in a manner that can understood by computer system. Face recognition systems are used on a wide range in mobile applications, banks, and military sites, police stations to make a recognition of the criminals, even in covid19 situations the face recognition was used too. Our proposed method aims to increase the accuracy a face recognition system using deep learning models, the system proposed is able to work on any platform compiled with python which is the application was used in our study. The model performed gives a high accuracy with a very speed result.

Detection of anomaly in surveillance videos using quantum convolutional neural networks

Anomalous behavior identification is the process of detecting behavior that differs from its normal. These incidents will vary from violence to war, road crashes to kidnapping, and so on in a surveillance model. Video anomaly detection from video surveillance is a difficult research activity due to the frequency of anomalous cases. Since certain devices need manual evaluation for the detection of violent or criminal situations at the same time video monitoring of security cameras is also a challenging task and is unreliable. When the data or model dimension is sufficiently large, convolutional neural networks have the limitation of learning inefficiently. Quantum Convolutional Neural Network (QCNN) is the name given to a technology that combines CNN and quantum computing. Quantum computation and CNN are combined to create a more efficient and outperforming solution for solving complicated machine-learning problems. To analyze the anomalies in a sequence of video frames, two models are proposed in this research. In this research 07 layers of Javeria deep convolutional neural network (DCNN) are proposed on the selected hyperparameters named J. DCNN which is also different from the existing models to analyze the abnormal behavior in a video segment. Furthermore, for a comprehensive analysis of the abnormal video frames a model is proposed which is the combination of Javeria quantum and convolutional neural networks (J. QCNN). In this model 04-qubit quantum neural network is used with five layers and an optimal loss rate named J. QCNN. The proposed J. QCNN model is different from the existing deep learning architectures. The proposed models are trained from the scratch for the detection of anomalous from top challenging publicly available video surveillance datasets such as UNI-Crime and UCF Crime. The proposed J. QCNN model classifies the number of violent robberies such as armed thefts containing handguns or knives, and robberies displaying varying levels of viciousness with 0.99 accuracy while J. DCNN model gives 0.97 accuracy. The obtained results are superior in comparison with recent existing cutting-edge published work for real-time anomaly detection in video CCTV.

Online Video Super-resolution using Information Replenishing Unidirectional Recurrent Model

Recurrent Neural Networks (RNN) are widespread for Video Super-Resolution (VSR) because of their proven ability to learn spatiotemporal inter-dependencies across the temporal dimension. Despite RNN’s ability to propagate memory across longer sequences of frames, vanishing gradient and error accumulation remain major obstacles to unidirectional RNNs in VSR. Several bi-directional recurrent models are suggested in the literature to alleviate this issue; however, these models are only applicable to offline use cases due to heavy demands for computational resources and the number of frames required per input. This paper proposes a novel unidirectional recurrent model for VSR, namely “Replenished Recurrency with Dual-Duct” (R2D2), that can be used in an online application setting. R2D2 incorporates a recurrent architecture with a sliding-window-based local alignment resulting in a recurrent hybrid architecture. It also uses a dual-duct residual network for concurrent and mutual refinement of local features along with global memory for full utilisation of the information available at each timestamp. With novel modelling and sophisticated optimisation, R2D2 demonstrates competitive performance and efficiency despite the lack of information available at each time-stamp compared to its offline (bi-directional) counterparts. Ablation analysis confirms the additive benefits of the proposed sub-components of R2D2 over baseline RNN models.The PyTorch-based code for the R2D2 model will be released at R2D2 GitRepo.

Molecular Characterization of Nine TRAF Genes in Yellow Catfish (Pelteobagrus fulvidraco) and Their Expression Profiling in Response to Edwardsiella ictaluri Infection.

The yellow catfish (Pelteobagrus fulvidraco) is an economic fish with a large breeding scale, and diseases have led to huge economic losses. Tumor necrosis factor receptor-associated factors (TRAFs) are a class of intracellular signal transduction proteins that play an important role in innate and adaptive immune responses by mediating NF-κB, JNK and MAPK signaling pathways. However, there are few studies on the TRAF gene family in yellow catfish. In this study, the open reading frame (ORF) sequences of TRAF1, TRAF2a, TRAF2b, TRAF3, TRAF4a, TRAF4b, TRAF5, TRAF6 and TRAF7 genes were cloned and identified in yellow catfish. The ORF sequences of the nine TRAF genes of yellow catfish (Pf_TRAF1-7) were 1413-2025 bp in length and encoded 470-674 amino acids. The predicted protein structures of Pf_TRAFs have typically conserved domains compared to mammals. The phylogenetic relationships showed that TRAF genes are conserved during evolution. Gene structure, motifs and syntenic analyses of TRAF genes showed that the exon-intron structure and conserved motifs of TRAF genes are diverse among seven vertebrate species, and the TRAF gene family is relatively conserved evolutionarily. Among them, TRAF1 is more closely related to TRAF2a and TRAF2b, and they may have evolved from a common ancestor. TRAF7 is quite different and distantly related to other TRAFs. Real-time quantitative PCR (qRT-PCR) results showed that all nine Pf_TRAF genes were constitutively expressed in 12 tissues of healthy yellow catfish, with higher mRNA expression levels in the gonad, spleen, brain and gill. After infection with Edwardsiella ictaluri, the expression levels of nine Pf_TRAF mRNAs were significantly changed in the head kidney, spleen, gill and brain tissues of yellow catfish, of which four genes were down-regulated and one gene was up-regulated in the head kidney; four genes were up-regulated and four genes were down-regulated in the spleen; two genes were down-regulated, one gene was up-regulated, and one gene was up-regulated and then down-regulated in the gill; one gene was up-regulated, one gene was down-regulated, and four genes were down-regulated and then up-regulated in the brain. These results indicate that Pf_TRAF genes might be involved in the immune response against bacterial infection. Subcellular localization results showed that all nine Pf_TRAFs were found localized in the cytoplasm, and Pf_TRAF2a, Pf_TRAF3 and Pf_TRAF4a could also be localized in the nucleus, uncovering that the subcellular localization of TRAF protein may be closely related to its structure and function in cellular mechanism. The results of this study suggest that the Pf_TRAF gene family plays important roles in the immune response against pathogen invasion and will provide basic information to further understand the roles of TRAF gene against bacterial infection in yellow catfish.

MSCDP: Multi-step crowd density predictor in indoor environment

Monitoring and predicting crowd movements in indoor environments are of great importance in crowd management to prevent crushing and trampling. Existing works mostly focused on individual trajectory forecasting in a less crowded scene, or crowd counting and density estimation. Only a very few works predict the crowd density distribution. However, this study is failing to realize multi-step prediction or exploits only density heatmaps modality and ignores the information complementation with corresponding video frames. Therefore, we are motivated to predict crowd density distribution in multiple time steps to facilitate long-term prediction. In this paper, a Multi-Step Crowd Density Predictor (MSCDP) to fuse video frame sequences and corresponding density heatmaps, is proposed to accurately forecast future crowd density heatmaps. To capture long-term periodic movement features, the long-term optical flow context memory (LOFCM) module is designed to store learnable patterns. We conducted extensive experiments on two real-world datasets. Evaluation results show that our MSCDP outperforms the state-of-the-art baseline techniques and MSCDP variants in terms of various prediction errors, demonstrating the effectiveness of MSCDP and each of its key components in multi-step crowd density prediction.

Large scale sequence alignment via efficient inference in generative models

Finding alignments between millions of reads and genome sequences is crucial in computational biology. Since the standard alignment algorithm has a large computational cost, heuristics have been developed to speed up this task. Though orders of magnitude faster, these methods lack theoretical guarantees and often have low sensitivity especially when reads have many insertions, deletions, and mismatches relative to the genome. Here we develop a theoretically principled and efficient algorithm that has high sensitivity across a wide range of insertion, deletion, and mutation rates. We frame sequence alignment as an inference problem in a probabilistic model. Given a reference database of reads and a query read, we find the match that maximizes a log-likelihood ratio of a reference read and query read being generated jointly from a probabilistic model versus independent models. The brute force solution to this problem computes joint and independent probabilities between each query and reference pair, and its complexity grows linearly with database size. We introduce a bucketing strategy where reads with higher log-likelihood ratio are mapped to the same bucket with high probability. Experimental results show that our method is more accurate than the state-of-the-art approaches in aligning long-reads from Pacific Bioscience sequencers to genome sequences.