Abstract: Provided is an efficient approach to establish a foundational video-text model for tasks including open-vocabulary video classification, text-to-video retrieval, video captioning and video question-answering. Some example implementations include a model which can be referred to as VideoCoCa. Example implementations reuse a pretrained image-text contrastive captioner (CoCa) model and adapt it to video-text tasks with little or minimal extra training. While previous works adapt image-text models with various cross-frame fusion modules (for example, cross-frame attention layer or perceiver resampler) and finetune the modified architecture on video-text data, aspects of the present disclosure leverage findings that the generative attentional pooling and contrastive attentional pooling layers in the image-text CoCa design are instantly adaptable to “flattened frame embeddings”, yielding a strong zero-shot transfer baseline for many video-text tasks.

Abstract: A video-specific super-resolution method includes obtaining an (i+1)th frame of image from a video, and obtaining image features of an ith frame of image in the video and long time series features before the ith frame of image, which are cached during super-resolution processing of the ith frame of image; performing super-resolution prediction on the image features of the ith frame of image, the long time series features before the ith frame of image, and the (i+1)th frame of image using a generative network, to obtain a super-resolution image of the (i+1)th frame of image, image features of the (i+1)th frame of image, and long time series features before the (i+1)th frame of image; and caching the image features of the (i+1)th frame of image and the long time series features before the (i+1)th frame of image, i being a positive integer greater than 2.

Abstract: Systems and methods are provided for vector-quantized image modeling using vision transformers and improved codebook handling. In particular, the present disclosure provides a Vector-quantized Image Modeling (VIM) approach that involves pre-training a machine learning model (e.g., Transformer model) to predict rasterized image tokens autoregressively. The discrete image tokens can be encoded from a learned Vision-Transformer-based VQGAN (example implementations of which can be referred to as ViT-VQGAN). The present disclosure proposes multiple improvements over vanilla VQGAN from architecture to codebook learning, yielding better efficiency and reconstruction fidelity. The improved ViT-VQGAN further improves vector-quantized image modeling tasks, including unconditional image generation, conditioned image generation (e.g., class-conditioned image generation), and unsupervised representation learning.

Abstract: A device and method for detecting an inter-turn electromagnetic pulse vibration wave characteristic of a turbogenerator rotor winding are provided. A signal source and a time sequence control circuit generate a high-potential abrupt electric field; circularly polarized electromagnetic waves generated by a parasitic inductive power supply and symmetrically deflecting by 180° are respectively coupled to a positive electrode and a negative electrode clockwise or counter-clockwise; a first turn on the positive electrode and a first turn on the negative electrode are mutually induced; as time goes by, energy is returned to the parasitic inductive power supply, and is sequentially conducted to a second turn; the parasitic inductive power supply and the second turn further start feeding back energy to the first turn in circular polarization; all turns sequentially perform feedback and superposition one another stage by stage; and all coupling turns show sinusoidal waves with a same time constant.

Abstract: A device and method for detecting an inter-turn electromagnetic pulse vibration wave characteristic of a turbogenerator rotor winding are provided. A signal source and a time sequence control circuit generate a high-potential abrupt electric field; circularly polarized electromagnetic waves generated by a parasitic inductive power supply and symmetrically deflecting by 180° are respectively coupled to a positive electrode and a negative electrode clockwise or counter-clockwise; a first turn on the positive electrode and a first turn on the negative electrode are mutually induced; as time goes by, energy is returned to the parasitic inductive power supply, and is sequentially conducted to a second turn; the parasitic inductive power supply and the second turn further start feeding back energy to the first turn in circular polarization; all turns sequentially perform feedback and superposition one another stage by stage; and all coupling turns show sinusoidal waves with a same time constant.

Abstract: A device and a method for nondestructively detecting a transient characteristic of a conductive screw of a turbo-generator rotor are provided. The device includes a personal computer (PC), an extremely-steep pulse generator, an ultra-high-frequency double-isolation transformer, and a pulse emitting and coupling module, which are connected in sequence. The pulse emitting and coupling module is connected to a load. A synchronous pulse receiving non-inductive divider circuit synchronously receives a characteristic waveform from the load, and the synchronous pulse receiving non-inductive divider circuit is connected to an ultra-high-speed analog/digital (A/D) module through a nonlinear saturation amplifying circuit that amplifies a signal. The PC receives a signal from the ultra-high-speed A/D module. The load includes a positive or negative excitation lead loop that is in a 180° symmetrical and instantaneous short-circuit state and a rotor shaft.

Abstract: Systems and methods are provided for vector-quantized image modeling using vision transformers and improved codebook handling. In particular, the present disclosure provides a Vector-quantized Image Modeling (VIM) approach that involves pretraining a machine learning model (e.g., Transformer model) to predict rasterized image tokens autoregressively. The discrete image tokens can be encoded from a learned Vision-Transformer-based VQGAN (example implementations of which can be referred to as ViT-VQGAN). The present disclosure proposes multiple improvements over vanilla VQGAN from architecture to codebook learning, yielding better efficiency and reconstruction fidelity. The improved ViT-VQGAN further improves vector-quantized image modeling tasks, including unconditional image generation, conditioned image generation (e.g., class-conditioned image generation), and unsupervised representation learning.

Abstract: An electronic device may include a display and an enclosure. The enclosure may include a housing, a front cover coupled to the housing and comprising a front cover member positioned over the display, and a rear cover coupled to the housing and including a rear cover member. The rear cover member may be formed from a glass material including metal nanoparticles configured to impart color to the glass material and having a dielectric constant from 5.5 to 7.5 in a frequency band from 5 GHz to 45 GHz. The rear cover member may include a first portion defining a first thickness and characterized by a first color, and a second portion defining a second thickness, greater than the first thickness, and characterized by a second color, different from the first color.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing multi-modal inputs using contrastive captioning neural networks.

Abstract: Example embodiments of the present disclosure relate to systems and methods for pretraining image-processing models on weakly-supervised image-text pairs. The pretraining can include receiving a training sequence for the machine-learned image-processing model. The training sequence can include text tokens and image tokens. A prefix sequence can contain the image tokens. A remainder sequence can include a remainder set of the text tokens. The pretraining can include determining, using the prefix sequence as an input to the machine-learned image-processing model, an objective based on recovery of the remainder sequence. The pretraining can include updating one or more learnable parameters of the machine-learned image-processing model based on the objective.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating labeled training data using a pre-trained language model neural network. In particular, the language model neural network can generate the text input in a new labeled training example from an input sequence that includes (i) one or more context inputs and (ii) a text label that identifies the ground truth category for the new labeled training example.

Abstract: A device and a method for nondestructively detecting a transient characteristic of a conductive screw of a turbo-generator rotor are provided. The device includes a personal computer (PC), an extremely-steep pulse generator, an ultra-high-frequency double-isolation transformer, and a pulse emitting and coupling module, which are connected in sequence. The pulse emitting and coupling module is connected to a load. A synchronous pulse receiving non-inductive divider circuit synchronously receives a characteristic waveform from the load, and the synchronous pulse receiving non-inductive divider circuit is connected to an ultra-high-speed analog/digital (A/D) module through a nonlinear saturation amplifying circuit that amplifies a signal. The PC receives a signal from the ultra-high-speed A/D module. The load includes a positive or negative excitation lead loop that is in a 180° symmetrical and instantaneous short-circuit state and a rotor shaft.

Abstract: The present application provides an intelligent device, an intelligent speaker, and a method and system for controlling the same. The intelligent device includes a first sound detection module configured to detect a first sound signal directly reaching the first sound detection module; an angle determination module configured to determine a time difference between the receiving time of the first sound signal and the receiving time of the second sound signal, and determine a relative angle between the intelligent device and the intelligent speaker based on a distance between the first sound detection module and the second sound detection module and the time difference; and a transmitting module configured to transmit a notification message containing the relative angle to the intelligent speaker, so that the intelligent speaker directionally transmits a sound to the intelligent device based on the relative angle. Directional sounding based on relative angle calculation is realized.

Abstract: The present disclosure discloses a maltooligosyl trehalose synthase mutant with improved thermal stability, and belongs to the technical fields of enzyme engineering and protein engineering. The residual enzyme activities of the MTSase mutants S361R, S444E, S361R/S444E, S361K/S444E, G415P/S361R/S444E and G415P consistent with the present disclosure after treatment at 60° C. for 10 min are respectively 70.3%, 50.1%, 83.5%, 65.9%, 100% and 80.7%, which are respectively 1.6, 1.1, 1.9, 1.5, 2.3 and 1.9 times of that of the wild type. The half-lives of the S361R/S444E and G415P/S361R/S444E at 60° C. are respectively 14.9 min and 90.8 min which are respectively 3.2 and 19.7 times of that of the wild type, indicating that the thermal stability of the MTSase mutant consistent with the present disclosure is significantly improved than that of the wild type.

Abstract: The present disclosure discloses a maltooligosyl trehalose synthase mutant with improved thermal stability, and belongs to the technical fields of enzyme engineering and protein engineering. The residual enzyme activities of the MTSase mutants S361R, S444E, S361R/S444E, S361K/S444E, G415P/S361R/S444E and G415P consistent with the present disclosure after treatment at 60° C. for 10 min are respectively 70.3%, 50.1%, 83.5%, 65.9%, 100% and 80.7%, which are respectively 1.6, 1.1, 1.9, 1.5, 2.3 and 1.9 times of that of the wild type. The half-lives of the S361R/S444E and G415P/S361R/S444E at 60° C. are respectively 14.9 min and 90.8 min which are respectively 3.2 and 19.7 times of that of the wild type, indicating that the thermal stability of the MTSase mutant consistent with the present disclosure is significantly improved than that of the wild type.