Abstract: Systems and methods are described for generating pixel image data, using a lensless camera, based on light that travels through a mask that with pattern masking the lensless camera. The system applies a transformation function to the pixel image data to generate frequency domain image data. The system inputs the frequency domain image data into a machine learning model, wherein the machine learning model does not have access to data that represents the pattern of the mask. The model is trained using a set of images with the feature that are captured by the flat, lensless camera through the mask. The system processes the frequency domain image data using the machine learning model to determine whether the pixel image data depicts the image feature. The system further performs an action based on determining that the pixel image data depicts the image feature.

Abstract: A system and methods for multi-type data compression or decompression with a virtual management layer in a distributed computing environment, comprising. It incorporates a virtual management layer to organize incoming data types and allocate compression or decompression tasks across multiple computing devices, selecting techniques best suited for particular data types. Associated data sets may be flagged prior to processing, ensuring preservation of relationships even when compressed or decompressed on different devices. This distributed approach allows efficient parallel processing of multiple data types, improving scalability and performance. A load balancing module optimizes task distribution based on available resources and processing requirements. The system enables each data type to be processed using the most efficient technique while maintaining associations between related data sets, effectively handling larger volumes of diverse data.

Abstract: The system trains a machine learning model using a loss function, with a part that penalizes overall signal loss, and a second part of the loss function that penalizes texture loss. The system computes a first neural feature of a first media frame stored by a media server using the trained machine learning model. The system causes a client device to receive a second media frame as a part of a media stream from the media server where the second frame is a modified version of the first media frame. The system causes the client to compute a second neural feature of the second media frame using the trained machine learning model, and compute a QoE metric based on the first neural feature and the second neural feature. The system receives the QoE metric, and uses it to modify at least one parameter of the media stream.

Abstract: A system and methods for upsampling compressed data using a jointly trained Vector Quantized Variational Autoencoder (VQ-VAE) and neural upsampler. The system compresses input data into a discrete latent space using a VQ-VAE encoder, reconstructs the data using a VQ-VAE decoder, and enhances the reconstructed data using a neural upsampler. The VQ-VAE and neural upsampler are jointly trained using a combined loss function, enabling end-to-end optimization. The system allows for efficient compression and high-quality reconstruction of various data types, including financial time-series, images, audio, video, sensor data, and text. The learned discrete latent space can be explored and manipulated using techniques such as interpolation, extrapolation, and vector arithmetic to generate new or modified data samples. The system finds applications in data storage, transmission, analysis, and generation across multiple domains.

Abstract: Systems and methods are provided for encoding a frame of 3D media content. The systems and methods may be configured to access a first frame of 3D media content and generate a data structure for the first frame based on color attributes information of the first frame, wherein each element of the data structure encodes a single color. The systems and methods may be configured to train a machine learning model based on the first frame of 3D media content, wherein the machine learning model is trained to receive as input a coordinate of a voxel of the first frame, and to output an identifier of a particular element in the generated data structure. The systems and methods may be configured to generate encoded data for the first frame based at least in part on weights of the trained machine learning model and the generated data structure.

Abstract: A system and methods for multi-type data compression or decompression with a virtual management layer, comprising. It incorporates a virtual management layer to organize incoming data types and select a compression or decompression system that utilizes a technique best suited for a particular data type. Associated data sets may be flagged prior to compression or decompression so that associated types may be preserved together after the compression or decompression process is complete. This approach allows each data type to be compressed or decompressed using a technique that is the most efficient for a particular data type. Additionally, the approach allows all information associated with a particular data set to be compressed or decompressed in some way.

Abstract: For compressing data, preprocessing operations are performed on raw input data. A discrete cosine transform is performed on the preprocessed data, and multiple subbands are created, where each subband represents a particular range of frequencies. The subbands are organized into multiple groups, where the multiple groups comprise a first low frequency group, a second low frequency group, and a high frequency group. A latent space representation is generated corresponding to each of the multiple groups of subbands. A first bitstream is created based on the latent space representation, and an alternate representation of the latent space is used for creating a second bitstream, enabling multiple-pass techniques for data compression. The multiple bitstreams may be multiplexed to form a combined bitstream for storage and/or transmission purposes.

Abstract: A system and methods for upsampling of decompressed genomic data after lossy compression using a neural network integrates AI-based techniques to enhance compression quality. It incorporates a novel deep-learning neural network that upsamples decompressed data to restore information lost during lossy compression, taking advantage of cross-correlations between genomic data sets.

Abstract: A ventilation method for a high gas working face based on alternating intake and air return in a mine gallery is provided. In this method, an isolated island working face is divided into several sections along a strike direction. A coal pillar is alternately set as a conventional section and a gas-drainage section. Opposite coal pillars in the two mine galleries within the same section are respectively used as the conventional section and the gas-drainage section. Two mine galleries on both sides of the isolated island working face are alternately used as an intake gallery and a return gallery. The coal pillar at the side of the mine gallery as the return gallery is the gas-drainage section. A gas-drainage hole communicating a goaf is provided in the gas-drainage section, so that gas in the goafs at two sides of the isolated island working face can be extracted alternately.

Abstract: A system and method for controllable lossy data compression employing a joint learning framework to efficiently compress and reconstruct input data while balancing compression ratio and reconstruction quality. The system comprises an encoding system, a temporal modeling system, and a decoding system, which are jointly optimized to minimize a combined loss function. The encoding system, such as a Vector Quantized Variational Autoencoder (VQ-VAE) compresses the input data into a compact representation, while introducing a controllable degree of lossy compression based on adjustable compression parameters. The temporal modeling system, such as a Multilayer Perceptron Long Short-Term Memory captures temporal dependencies in the compressed representation. The decoding system, such as a VQ-VAE decoder, reconstructs the input data from the compressed representation. By providing control over the trade-off between compression ratio and reconstruction quality, the system offers flexibility for diverse applications.

Abstract: A system and method for real time discrete cosine transform image and video processing with convolutional neural network architecture. The system and method incorporate discrete cosine transform image processing with convolutional neural networks to achieve fast and efficient image processing that yields more reliable results than previously used image processing methods. The proposed system and method enable effective, real time, image processing which is applicable to a wide range of imaging and video devices.

Abstract: Systems and methods are provided for efficiently encoding geometry information for 3D media content. An illustrative system generates a low-resolution polygon mesh from a high-resolution polygon mesh. The system uses a vertex occupancy prediction network to generate, from vertices of the low-resolution polygon mesh, approximated vertices of the high-resolution polygon mesh. The system uses a connectivity prediction network to generate, from approximated vertices of the high-resolution polygon mesh, approximated connections of the high-resolution polygon mesh. The system computes vertex errors between the approximated vertices and the vertices of the high-resolution polygon mesh, and connectivity errors between the approximated connections and the connections of the high-resolution polygon mesh.

Abstract: A system and method are provided for encoding frames of 3D media content. The system trains first and second neural networks based on the voxel geometry information of their respective frames, such that the neural network is configured to receive a coordinate of a voxel and output color attributes information for the voxel. The trained neural network comprises a plurality of weights for each layer of the neural network. The system generates encoding data for the first frame by storing each respective decomposition data of a first plurality of matrices generated based on the plurality of weights for each layer of the first neural network. The system generates encoding data for the second frame by storing differences between the decomposition data of the second plurality of matrices and the decomposition data of the first plurality of matrices.

Abstract: A system and method are disclosed for generating hyperspectral images from RGB (red-green-blue) images. A set of data includes training hyperspectral images and their corresponding RGB images. A spectral band grouping is performed on the training hyperspectral images based on a correlation coefficient of spectral bands. A decomposition network is used to generate a reconstructed hyperspectral image. A fine-tuning network is used to create a reconstructed RGB images. The difference between an input RGB image and a corresponding reconstructed RGB image is used to adjust one or more weights of one or more of the networks.

Abstract: The system trains a machine learning model using a loss function, with a part that penalizes overall signal loss, and a second part of the loss function that penalizes texture loss. The system computes a first neural feature of a first media frame stored by a media server using the trained machine learning model. The system causes a client device to receive a second media frame as a part of a media stream from the media server where the second frame is a modified version of the first media frame. The system causes the client to compute a second neural feature of the second media frame using the trained machine learning model, and compute a QoE metric based on the first neural feature and the second neural feature. The system receives the QoE metric, and uses it to modify at least one parameter of the media stream.

Abstract: A ventilation method for a high gas working face based on alternating intake and air return in a mine gallery is provided. In this method, an isolated island working face is divided into several sections along a strike direction. A coal pillar is alternately set as a conventional section and a gas-drainage section. Opposite coal pillars in the two mine galleries within the same section are respectively used as the conventional section and the gas-drainage section. Two mine galleries on both sides of the isolated island working face are alternately used as an intake gallery and a return gallery. The coal pillar at the side of the mine gallery as the return gallery is the gas-drainage section. A gas-drainage hole communicating a goaf is provided in the gas-drainage section, so that gas in the goafs at two sides of the isolated island working face can be extracted alternately.

Abstract: A system and method are disclosed for low-light image enhancement using denoising preprocessing with wavelet decomposition AI-based techniques to enhance image quality of low-light images. Subsampled images are created from a raw input image. A wavelet decomposition process is performed on each subimage to create multiple frequency domain subimages. Each frequency domain subimage is input into a corresponding neural network. The output of each corresponding network is input to an inverse wavelet module. The output of the inverse wavelet module is a denoised image that is input to an image signal processing pipeline, where additional processing may be performed on the denoised image.

Abstract: A system and methods for multi-type data compression or decompression with a virtual management layer, comprising. It incorporates a virtual management layer to organize incoming data types and select a compression or decompression system that utilizes a technique best suited for a particular data type. Associated data sets may be flagged prior to compression or decompression so that associated types may be preserved together after the compression or decompression process is complete. This approach allows each data type to be compressed or decompressed using a technique that is the most efficient for a particular data type. Additionally, the approach allows all information associated with a particular data set to be compressed or decompressed in some way.

Abstract: A diffractive optical waveguide includes: an optical waveguide substrate, at least one coupling-in grating, at least one relay grating, and at least one coupling-out grating. The coupling-in grating is arranged in the coupling-in region. The relay grating is arranged in the relay region. The coupling-out grating is arranged in the coupling-out region. The relay grating and the coupling-out grating are both two-dimensional gratings. The coupling-in region and the coupling-out region are both symmetrical in shape, the relay grating is symmetrical or asymmetrical in shape. Thus, the relay grating can change a transmission path of part of lights, the lights coupled-in from the coupling-in grating can be transmitted to the coupling-out region through the relay grating and cover the entire coupling-out region, avoiding the problem of dark angle of fields of view in the coupling-out region and improving energy utilization efficiency of the lights.