Abstract: A semantic segmentation network model uncertainty quantification method based on evidence inference. The method comprises the steps of constructing an FCN network model, and training the FCN network model by using a training data set to obtain a trained FCN network model for semantic segmentation of image data; transplanting a D-S theory of evidence to the trained FCN network model to obtain a reconstructed FCN network model; and inputting to-be-segmented image data into the reconstructed FCN network model, outputting a classification result of a to-be-segmented image by the FCN network model, and calculating a classification result uncertainty value of each pixel point by using the D-S theory of evidence index.

Abstract: Certain aspects of the present disclosure provide techniques for performing machine learning computations in a compute in memory (CIM) array comprising a plurality of bit cells, including: determining that a sparsity of input data to a machine learning model exceeds an input data sparsity threshold; disabling one or more bit cells in the CIM array based on the sparsity of the input data prior to processing the input data; processing the input data with bit cells not disabled in the CIM array to generate an output value; applying a compensation to the output value based on the sparsity to generate a compensated output value; and outputting the compensated output value.

Abstract: The present invention provides techniques for learning Mahalanobis distance similarity metrics from data for individually fair machine learning models. In one aspect, a method for learning a fair Mahalanobis distance similarity metric includes: obtaining data with similarity annotations; selecting, based on the data obtained, a model for learning a Mahalanobis covariance matrix ?; and learning the Mahalanobis covariance matrix ? from the data using the model selected, wherein the Mahalanobis covariance matrix ? fully defines the fair Mahalanobis distance similarity metric.

Abstract: A system and method for classification of voice samples to genuine voice samples or spoofing voice samples may include: extracting a set of features from each of a plurality of voice samples, each voice sample labeled as genuine or spoof; training a neural network having a plurality of nodes organized into layers, with links between the nodes, wherein each link comprises a weight, with the sets of features, by adjusting at least one of the weights using a loss function that comprises a regulation factor, wherein the regulation factor is set to zero for voice samples labeled as genuine and is proportional to the prediction of the neural network for data samples labeled as spoofing.

Abstract: Present disclosure relates to management of artificial intelligence systems by identifying root cause of reduced performance and/or failure in computing systems, and particularly relates to systems and methods for detecting a drift in supervised and unsupervised machine learning (ML) models. The system retrieves current dataset corresponding to output of supervised ML models and unsupervised ML models. Further, the system segregates the current dataset based on requirement of a drift detection model and applies a plurality of drift detection models to the segregated dataset to generate predictive results corresponding to the current dataset. Furthermore, the system determines errors in predictive results by comparing predictive results to reference values associated with current dataset. Additionally, the system detects the drift in supervised ML models and unsupervised ML models based on determined errors being above a threshold value.

Abstract: A system performs operations that include receiving, via first computing environment, a request to process text data using a first natural language processing (NLP) model. The operations further include accessing configuration data associated with the NLP model, where the configuration data generated using a domain specific language that supports a plurality of preprocessing modules in a plurality of programming languages. The operations also include selecting, based on the configuration data, one or more preprocessing modules of the plurality of preprocessing modules, generating, based on the configuration data, a preprocessing pipeline using the one or more preprocessing modules, and generating preprocessed text data by inputting the text data into the preprocessing pipeline. The preprocessed text data is provided to the first NLP model.

Abstract: The invention relates to the provision of apparatus and a method to determine the operation of a vehicle and provide monitoring of the same and the condition of a storage area capable of being transported as part of the vehicle. The storage area includes one or more detection devices provided in the same and which are in communication with the operator of the vehicle and/or external personnel or organisations so as to detect a change in condition of the vehicle and in particular if the change exceeds predetermined parameters.

Abstract: Certain aspects involve a machine-learning query system that uses a dual deep learning network to service queries and other requests. In one example, a machine-learning query system services a query received from a client computing system. A dual deep learning network included in the machine-learning query system matches an unstructured input data object, received from the client computing system, to an unstructured reference data object. The matching may include generating an input feature vector by an embedding subnetwork, based on the unstructured input data object. The matching may also include generating an output probability by a relationship subnetwork, based on the input feature vector and a relationship feature vector that is based on the unstructured reference data object. The machine-learning query system may transmit a responsive message to the client system.

Abstract: A compute engine is described. The compute engine includes compute-in-memory (CIM) modules and may include an input buffer coupled with the CIM modules. The input buffer stores a vector. The CIM modules store weights corresponding to a matrix and perform a vector-matrix multiplication (VMM) for the matrix and the vector. The CIM modules further include storage cells and vector multiplication units (VMUs) coupled with the storage cell and, if present, the input buffer. The storage cells store the weights. The VMUs multiply, with the vector, at least a portion of each weight of a portion of the plurality of weights corresponding to a portion of the matrix. A set of VMUs performs multiplications for a first weight length and a second weight length different from the first weight length such that each VMU of the set performs multiplications for both the first weight length and the second weight length.

Abstract: The present application discloses a computing system for implementing an artificial neural network model. The artificial neural network model has a structure of multiple layers. The computing system comprises a first processing unit, a second processing unit, and a third processing unit. The first processing unit performs computations of the first layer based on a first part of input data of the first layer to generate a first part of output data. The second processing unit performs computations of the first layer based on a second part of the input data of the first layer so as to generate a second part of the output data. The third processing unit performs computations of the second layer based on the first part and the second part of the output data. The first processing unit, the second processing unit, and the third processing unit have the same structure.

Abstract: Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a base station may transmit, to a user equipment (UE), a federated learning configuration that indicates one or more parameters of a federated learning procedure associated with a machine learning component. The base station may receive a local update associated with the machine learning component from the UE based at least in part on the federated learning configuration. Numerous other aspects are provided.

Abstract: An image generation device derives, for a subject including a specific structure, a subject model representing the subject by deriving each feature amount of the target image having the at least one representation format and combining the feature amounts based on the target image. A latent variable derivation unit derives a latent variable obtained by dimensionally compressing a feature of the subject model according to the target information based on the target information and the subject model. A virtual image derivation unit outputs a virtual image having the representation format represented by the target information based on the target information, the subject model, and the latent variable.

Abstract: A method comprising: receiving a first workload data set, the first workload data set specifying a cache hit outcome distribution that is associated with a plurality of input-output (I/O) operations; identifying a plurality of workload portions of the first workload data set, each of the workload portions identifying: (i) a rate of a cache hit outcome that is associated with a respective I/O operation, and (ii) a data size that is associated with the respective I/O operations; generating a plurality of initial vectors, each of the initial vectors being generated based on a different one of the plurality of workload portions, each of the initial vectors being generated by a different sub-network of a correlation neural network; generating a context vector based on the plurality of initial vectors; processing the context vector with a decoder to generate a plurality of data points in a response curve of a storage system.

Abstract: Various disclosed embodiments are directed to using one or more algorithms or models to select a suitable or optimal variation, among multiple variations, of a given content item based on feedback. Such feedback guides the algorithm or model to arrive at suitable variation result such that the variation result is produced as the output for consumption by users. Further, various embodiments resolve tedious manual user input requirements and reduce computing resource consumption, among other things, as described in more detail below.

Abstract: A method for ascertaining an uncertainty of a prediction of a first machine learning system. The method includes: processing a detected input variable by way of the first machine learning system, intermediate results, which are ascertained during the processing of the input variable by way of the machine learning system, being stored. Processing at least one of the stored intermediate results by way of a second machine learning system, the second machine learning system outputting an output variable, which characterizes an uncertainty of the output classification. A method for training the second machine learning system and a computer system, computer program, and a machine-readable memory element, on which the computer program is stored, are also described.

Abstract: The present invention relates to the technical field of network security, and in particular to an abnormal data detection method, system and device for industrial Internet. This detection method compares data distribution of an initial node with a normal feature expression performance in first normal data distribution subject to extraction processing to obtain a first anomaly score, compares the data distribution of the initial node with the normal feature expression performance in second normal data distribution subject to enhancement processing to obtain a second anomaly score, obtains a risk level of the node based on the first anomaly score and the second anomaly score, and immediately provides corresponding limits on a node communication permission; and the method provides dual detection, is high in accuracy and stable in detection results, and facilitates the maintenance of industrial Internet security.

Abstract: Methods and system are provided for tuning an artificial intelligence (“AI”) system. The methods and system may include training the AI system to identify distinct users their distinct interactions. The training may include receiving datatsets related to a plurality of distinct users and their plurality of distinct interactions. The AI system may parse the datasets to arrive at a plurality of sets of distinct interactions performed by respective distinct users, each distinct interaction may be associated with a level of security clearance. The AI system may receive a request from a user to access information. The methods and system may include the AI system initiating a communication with the user. The communication may be compared with the each set of distinct interactions until a user and level of security can be identified. The AI system may prompt the user to access information on the level of security clearance identified.

Abstract: A vision transformer network having extremely low latency and usable on mobile devices, such as smart eyewear devices and other augmented reality (AR) and virtual reality (VR) devices. The transformer network processes an input image, and the network includes a convolution stem configured to patch embed the image. A first stack of stages including at least two stages of 4-Dimension (4D) metablocks (MBs) (MB4D) follow the convolution stem. A second stack of stages including at least two stages of 3-Dimension MBs (MB3D) follow the MB4D stages. Each of the MB4D stages and each of the MB3D stages include different layer configurations, and each of the MB4D stages and each of the MB3D stages include a token mixer. The MB3D stages each additionally include a multi-head self attention (MHSA) processing block.

Abstract: A neural processor includes neural engines for performing convolution operations on input data corresponding to one or more tasks to generate output data. The neural processor also includes a data processor circuit coupled to external system memory. The data processor circuit includes a buffer for storing the output data from the neural engines. The neural processor further includes a task manager coupled to the data processor circuit. The task manager receives a context-switch task. The context-switch task specifies a switch of the data processor circuit from handling an outgoing task to an incoming task. The task manager sends configuration data of the context-switch task to cause the data processor circuit to transmit the output data corresponding to the outgoing task from the buffer to the external system memory. The data processor circuit also fetches data corresponding to the incoming task from the external system memory to the buffer.

Abstract: A block-based inference method for a memory-efficient convolutional neural network implementation is performed to process an input image. A block-based inference step is performed to execute a multi-layer convolution operation on each of a plurality of input block data to generate an output block data and includes selecting a plurality of ith layer recomputing features according to a position of the output block data along a scanning line feed direction, and then selecting an ith layer recomputing input feature block data according to the position of the output block data and the ith layer recomputing features, and selecting a plurality of ith layer reusing features according to the ith layer recomputing input feature block data along a block scanning direction, and then combining the ith layer recomputing input feature block data with the ith layer reusing features to generate an ith layer reusing input feature block data.

Abstract: Provided is a smart surveillance system that includes one unmanned aerial vehicle (UAV), a plurality of Internet of Things (IoT) terminals distributed in a surveillance area, and a plurality of base stations distributed in the surveillance area, wherein the UAV selects any IoT terminal from among the plurality of IoT terminals using deep learning auction training, receives surveillance data from the selected IoT terminal, and transmits the surveillance data to a data center through the Internet and any one base station among the plurality of base stations.

Abstract: Computer implemented methods, systems, and computer program products include program code executing on a processor(s) that obtains text from a computing resources in a domain and segments it into sentences. The processor(s) identifies entities in each sentence. The processor(s) classifies each entity with a sentiment (a polarity). The processor(s) identifies, for each sentence, a given entity with a strongest sentiment. The processor(s) derives for each sentence, for each entity, emotions and classifies each emotion by assigning a polarity to each and scoring each. The processor(s) calculates, based on the (emotion) scores, a mean and a standard deviation from the mean. The processor(s) maps the given entity for each sentence to a strongest emotion for the given entity where the polarity of the strongest emotion is the same as the polarity of the strongest sentiment. The processor(s) determines if each strongest emotion is within the standard deviation from the mean.

Abstract: An activation function conversion program unit and method may be configured to approximate a target activation function to a programmed activation function through machine-learning of an artificial neural network. The method may include setting up a target activation function; approximating the target activation function to a programmed activation function by machine-learning an artificial neural network; and converting the programmed activation function into a slope and offset and storing it in a lookup table. Accordingly, the computation speed and power consumption of the programmed activation function execution unit of an NPU may be optimized.

Abstract: Example devices, systems, and techniques predict renal denervation efficacy for reducing hypertension in a patient based on pulse information. For example, a system may include processing circuitry configured to obtain pulse information representative of pulses from both wrists of a patient, obtain a plurality of values representative of respective patient metrics for the patient, and apply the pulse information and the plurality of values to a deep learning model trained to represent a relationship of the pulse information and the patient metrics to an efficacy of renal denervation in reducing hypertension. In some examples, responsive to applying the pulse information and the plurality of values to the deep learning model, the processing circuitry obtains, from the deep learning model, a score indicative of renal denervation efficacy in reducing hypertension for the patient, and generates a graphical user interface comprising a graphical representation of the score for the patient.

Abstract: Systems, devices and methods for optimising and managing distributed energy storage and flexibility resources on a localised and group aggregation basis, particularly around the determination, analysis and predictive learning of local data patterns, scoring availability for flexibility and risk profiles, to inform the optimisation of energy supply and behind the meter storage resources and local clusters of co-located or close resources within a community, low voltage network, feeder, neighbourhood or building. Said optimisation to involve scheduled, reactive and active management of data sources and local clusters of resources, for a range of goals such as price, energy supply, renewable leverage, asset value, constraint or risk management.

Abstract: A method includes obtaining a first dataset associated with a first set of classes and a second dataset associated with a second set of classes, where the first and second datasets share features. The method also includes creating an aggregated dataset having a quality score for the second set of classes that satisfy a first threshold. Creating the aggregated dataset includes determining a first and second set of clusters based on the first and second datasets. The method includes determining a distance based on a record of the first set of clusters and a second record of the second set of clusters in a feature space. The method further includes determining a set of analysis scores based on the distance and generating an indication for a set of records associated with the set of analysis scores based on the set of analysis scores satisfying a second threshold.

Abstract: Systems and methods for using a user simulation model to facilitate detection of usage anomalies is disclosed. Usage data is received for a session between a user device and an application or service. The usage data is monitored to detect a usage anomaly, such as unusual or suspicious transactions, unexpected user or device attributes, or abnormal usage patterns. In response to detecting a request to terminate the session, the session is instead handed off to a user simulation model that simulates interactions of a user in the session. The user simulation model can be a machine learning model that is trained, using a training dataset, to simulate user interactions. When the user logs into the application or service, the session can be handed off from the user simulation model to the user, such that the session is perpetual or substantially perpetual.

Abstract: A method includes, for each floating-point layer in a set of floating-point layers: calculating a set of input activations and a set of output activations of the floating-point layer; converting the floating-point layer to a low-bit-width layer; calculating a set of low-bit-width output activations based on the set of input activations; and calculating a per-layer deviation statistic of the low-bit-width layer. The method also includes ordering the set of low-bit-width layers based on the per-layer deviation statistic of each low-bit-width layer.

Abstract: Techniques in placement of compute and memory for accelerated deep learning provide improvements in one or more of accuracy, performance, and energy efficiency. An array of processing elements comprising a portion of a neural network accelerator performs flow-based computations on wavelets of data. Each processing element comprises a compute element to execute programmed instructions using the data and a router to route the wavelets. The routing is in accordance with virtual channel specifiers of the wavelets and controlled by routing configuration information of the router. A software stack determines placement of compute resources and memory resources based on a description of a neural network. The determined placement is used to configure the routers including usage of the respective colors. The determined placement is used to configure the compute elements including the respective programmed instructions each is configured to execute.

Abstract: Methods and systems for language processing include augmenting an original training dataset to produce an augmented dataset that includes a first example that includes a first scrambled replacement for a first word and a definition of the first word, and a second example that includes a second scrambled replacement for the first word and a definition of an alternative to the first word. A neural network classifier is trained using the augmented dataset.

Abstract: The embodiments disclose a method including receiving, with a first memory device, user profile data including demographics from at least from two persons, receiving specimen samples of personal tissue and pelvic fluid from the at least two persons, analyzing, with a first processor, the specimen samples to identify test results including hormones, infectious diseases, PH levels and genetic disorders, comparing, with a second processor, the test results to determine any correlation information with predetermined medical conditions data, converting, with a third processor, the correlation information into recommendation data for each person, and transmitting, with a platform communication device, the recommendation data to each person via a respective wellness app operating on a mobile device of each person.

Abstract: Learning algorithms for oscillatory memristive neuromorphic circuits are described. In one embodiment, a neuromorphic circuit learning network includes a number of neuromorphic circuit nodes, each including a recognition neuron unit and a generative neuron unit. The learning network further includes a plurality of neuromorphic circuit feedforward couplings between the recognition neuron units in the neuromorphic circuit nodes, and a plurality of neuromorphic circuit feedback couplings between the generative neuron units in the neuromorphic circuit nodes. The learning network also includes a learning controller configured to drive activity among the recognition neuron units and train the generative neuron units for learning in one mode and to drive activity among the generative neuron units and train the recognition neuron units for learning in another mode. Various deep learning algorithms can be implemented in the learning network.

Abstract: A data processing system and method are disclosed, for implementing a windowed operation in at least three traversed dimensions. The data processing system maps the windowed operation in at least three traversed dimensions to a plurality of constituent windowed operations in two traversed dimensions. This plurality of 2-D windowed operations is implemented as such in at least one hardware accelerator. The data processing system assembles the results of the constituent 2-D windowed operations to produce the result of the windowed operation in at least three traversed dimensions.

Abstract: Systems and methods for multiple inertial measurement unit sensor fusion using machine learning are provided herein. In certain embodiments, a system includes inertial sensors that produce inertial measurements, a memory unit that stores a fusion model produced by at least one machine learning algorithm, and a processor that receives inertial measurements, where the processor applies the fusion model to the inertial measurements. The fusion model directs the processor to extract features from the inertial measurements, and to select inertial measurements based on a sensor in the plurality of inertial sensors that produced the inertial measurements. Also, the fusion model directs the processor to apply weights to the selected inertial measurements based on the extracted features, to apply compensation coefficients to the selected inertial measurements, and to fuse the selected inertial measurements into an inertial navigation solution.

Abstract: A method optimizes machine learning systems. A computing device accesses a committee of classifiers that have been trained using an initial labeled instance of data from an annotator. The initial labeled instance of data includes annotator-ranked attributes of the data, initial values of the attributes, and an initial prediction label that describes an initial predicted state based on the values. The computing system compares the attributes ranking from the annotator to attributes rankings that are generated by and used by each of the machine learning systems when evaluating one or more instances of unlabeled data that include the attributes, and weights the machine learning systems according to how closely each of the attributes rankings generated by and used by each of the machine learning systems match the attributes ranking from the annotator. The machine learning systems are then optimized based on this matching.

Abstract: An electronic apparatus is provided.

Abstract: This disclosure relates to techniques for executing dialectical analyses using large language models and/or other types of deep learning models. A dialectic logic engine can execute various programmatic processes or functions associated with applying dialectic analyses to input strings. The programmatic processes executed by the dialectic logic engine can initiate communication exchanges with one or more generative language models to derive parameters for performing dialectic analyses and/or to derive outputs based on the parameters. The dialectic logic engine also can execute functions for enforcing constraint conditions and/or eliminating bias from responses generated by the one or more generative language models to improve the accuracy, precision, and quality of the parameters and/or outputs derived from the parameters.

Abstract: An Online Machine Learning System (OMLS) includes an Online Machine Learning Engine (OMLE) for incorporating and utilizing one or more machine learning algorithms or models utilizing features to generate a result, and capable of incorporating and utilizing multiple different machine learning algorithms; wherein the OMLS is configured to perform continuous online machine learning, the continuous online machine learning comprising: continuous online machine learning from streaming data including an instance comprising a vector of inputs, the vector of inputs comprising a plurality of continuous or categorical features; and continuous online machine learning from periodically provided expert feedback.

Abstract: A method for operating an industrial automation system may involve receiving, via a first module of a plurality of modules in a control system, an indication that an error between a measurement associated with a target variable that corresponds with at least a portion of the industrial automation system and a modeled value for the target variable. The method may then involve determining, via the first module, whether the error is within a first range of values and retraining a model used to generate the modeled value for the target variable based on a portion of a plurality of sets of data points acquired via a plurality of sensors disposed in the industrial automation system in response to the error being within the first range of values.

Abstract: Hierarchical methods for selecting fixed point number formats with reduced mantissa bit lengths for representing values input to, and/or output, from, the layers of a DNN. The methods begin with one or more initial fixed point number formats for each layer. The layers are divided into subsets of layers and the mantissa bit lengths of the fixed point number formats are iteratively reduced from the initial fixed point number formats on a per subset basis. If a reduction causes the output error of the DNN to exceed an error threshold, then the reduction is discarded, and no more reductions are made to the layers of the subset. Otherwise a further reduction is made to the fixed point number formats for the layers in that subset. Once no further reductions can be made to any of the subsets the method is repeated for continually increasing numbers of subsets until a predetermined number of layers per subset is achieved.

Abstract: A computer-implemented method includes generating, based on a representation of a tensor mapping between an input tensor and an output tensor, a list of mappings from elements of the input tensor to elements of the output tensor, and generating groups of mappings from the list of mappings, where each of the groups of mappings corresponds to a respective set of matrix multiplications, a matrix transpose, or both. The computer-implemented method also includes generating a respective expression for each of the groups of mappings and generating code for summing results of the respective expressions, where each respective expression includes the respective set of matrix multiplications, the matrix transpose, or both.

Abstract: A machine learning system that uses a split net configuration to incorporate arbitrary constraints receives a set of input data and a set of functional constraints. The machine learning system jointly optimizes a deep learning model by using the set of input data and a wide learning model by using the set of constraints. The deep learning model includes an input layer, an output layer, and an intermediate layer between the input layer and the output layer. The wide learning model includes an input layer and an output layer but no intermediate layer. The machine learning system provides a machine learning model comprising the optimized deep learning model and the optimized wide learning model.

Abstract: Provided is a method and system with deep learning model generation. The method includes identifying a plurality of connections in a neural network that is pre-associated with a deep learning model, generating a plurality of pruned neural networks by pruning different sets of one or more of the plurality of connections to respectively generate each of the plurality of pruned neural networks, generating a plurality of intermediate deep learning models by generating a respective intermediate deep learning model corresponding to each of the plurality of pruned neural networks, and selecting one of the plurality of intermediate deep learning models, having a determined greatest accuracy among the plurality of intermediate deep learning models, to be an optimized deep learning model.

Abstract: A generative neural network control system controls a generative neural network by modifying the intermediate latent space in the generative neural network. The generative neural network includes multiple layers each generating a set of activation values. An initial layer (and optionally additional layers) receives an input latent vector, and a final layer outputs an image generated based on the input latent vector. The data that is input to each layer (other than the initial layer) is referred to as data in an intermediate latent space. The data in the intermediate latent space includes activation values (e.g., generated by the previous layer or modified using various techniques) and optionally a latent vector. The generative neural network control system modifies the intermediate latent space to achieve various different effects when generating a new image.

Abstract: Techniques that facilitate machine learning using multi-dimensional time series data are provided. In one example, a system includes a snapshot component and a machine learning component. The snapshot component generates a first sequence of multi-dimensional time series data and a second sequence of multi-dimensional time series data from multi-dimensional time series data associated with at least two different data types generated by a data system over a consecutive period of time. The machine learning component that analyzes the first sequence of multi-dimensional time series data and the second sequence of multi-dimensional time series data using a convolutional neural network system to predict an event associated with the multi-dimensional time series data.

Abstract: A control device executes the following processes: a first process of detecting that autonomous traveling control is not able to be continued and executing stop control of a vehicle; a second process of determining whether a person is present in the vehicle; and a third process of unlocking a door lock of the vehicle when at least one person is present in the vehicle after the vehicle is stopped by the first process.

Abstract: An example operation includes one or more of determining a sensor on a transport is not functioning properly, determining a severity of the malfunction, responsive to the severity exceeding a threshold, lowering an autonomous level of the transport, and responsive to the severity continuing to exceed the threshold, limiting an operation of the transport based on an intended output of the malfunctioning sensor.

Abstract: In various examples, a neural network may be trained for use in vehicle re-identification tasks—e.g., matching appearances and classifications of vehicles across frames—in a camera network. The neural network may be trained to learn an embedding space such that embeddings corresponding to vehicles of the same identify are projected closer to one another within the embedding space, as compared to vehicles representing different identities. To accurately and efficiently learn the embedding space, the neural network may be trained using a contrastive loss function or a triplet loss function. In addition, to further improve accuracy and efficiency, a sampling technique—referred to herein as batch sample—may be used to identify embeddings, during training, that are most meaningful for updating parameters of the neural network.

Abstract: A method for providing an intelligent electronic communication to a consumer device using a recommendation engine.

Abstract: With rapidly evolving technologies and emerging tools, sports-related videos generated online are rapidly increasing. To automate the sports video editing/highlight generation process, a key task is to precisely recognize and locate events-of-interest in videos. Embodiments herein comprise a two-stage paradigm to detect categories of events and when these events happen in videos. In one or more embodiments, multiple action recognition models extract high-level semantic features, and a transformer-based temporal detection module locates target events. These novel approaches achieved state-of-the-art performance in both action spotting and replay grounding. While presented in the context of sports, it shall be noted that the systems and methods herein may be used for videos comprising other content and events.