Abstract: Deep neural networks (DNN) model quantization may be used to reduce storage and computation burdens by decreasing the bit width. Presented herein are novel cursor-based adaptive quantization embodiments. In embodiments, a multiple bits quantization mechanism is formulated as a differentiable architecture search (DAS) process with a continuous cursor that represents a possible quantization bit. In embodiments, the cursor-based DAS adaptively searches for a quantization bit for each layer. The DAS process may be accelerated via an alternative approximate optimization process, which is designed for mixed quantization scheme of a DNN model. In embodiments, a new loss function is used in the search process to simultaneously optimize accuracy and parameter size of the model. In a quantization step, the closest two integers to the cursor may be adopted as the bits to quantize the DNN together to reduce the quantization noise and avoid the local convergence problem.

Abstract: A method and an apparatus for measuring temperature, and a computer-readable storage medium includes detecting a target position of an object in an input image; determining key points of the target position and weight information of each key point based on a detection result of the target position, in which the weight information is configured to indicate a probability of each key point being covered; acquiring temperature information of each key point; and determining a temperature of the target position at least based on the temperature information and the weight information of each key point.

Abstract: A method for training a face recognition model includes: acquiring a plurality of first training images being uncovered face images, and acquiring a plurality of covering object images; generating a plurality of second training images by separately fusing the plurality of covering object images with the uncovered face images; and training the face recognition model by inputting the plurality of first training images and the plurality of second training images into the face recognition model.

Abstract: Network architecture search (NAS) received a lot of attention. The supernet-based differentiable approach is popular because it can effectively share the weights and lead to more efficient search. However, the mismatch between the architecture and weights caused by weight sharing still exists. Moreover, the coupling effects among different operators are also neglected. To alleviate these problems, embodiments of an effective NAS methodology by similarity-based operator ranking are presented herein. With the aim of approximating each layer's output in the supernet, a similarity-based operator ranking based on statistical random comparison is used. In one or more embodiments, then the operator that possibly causes the least change to feature distribution discrepancy is pruned. In one or more embodiments, a fair sampling process may be used to mitigate the operators' Matthew effect that happened frequently in previous supernet approaches.

Abstract: An information processing method, a device, a system, and a storage medium. The information processing method includes: an AI camera first obtains real-time data in a unmanned retail scenario and performs a front-end processing on the real-time data based on a neural network model, where the front-end processing includes any one or more of commodity identifying and human body monitoring, and then transmits a result of the front-end processing to a server, where the result of the front-end processing is used to trigger the server to perform face recognition and/or determine a flow direction of a commodity according to the result of the front-end processing. The cost of the entire unmanned retail distributed system and the pressure on data transmission bandwidth can be reduced, and system scalability as well as the performance of the solution to the unmanned retail can be improved effectively.

Abstract: Tensor decomposition can be advantageous for compressing deep neural networks (DNNs). In many applications of DNNs, reducing the number of parameters and computation workload is helpful to accelerate inference speed in deployment. Modern DNNs comprise multiple layers with multi-array weights where tensor decomposition is a natural way to perform compression—in which the weight tensors in convolutional layers or fully-connected layers are decomposed with specified tensor ranks (e.g., canonical ranks, tensor train ranks). Conventional tensor decomposition with DNNs involves selecting ranks manually, which requires tedious human efforts to finetune the performance. Accordingly, presented herein are rank selection embodiments, which are inspired by reinforcement learning, to automatically select ranks in tensor decomposition.

Abstract: Deep neural networks (DNN) model quantization may be used to reduce storage and computation burdens by decreasing the bit width. Presented herein are novel cursor-based adaptive quantization embodiments. In embodiments, a multiple bits quantization mechanism is formulated as a differentiable architecture search (DAS) process with a continuous cursor that represents a possible quantization bit. In embodiments, the cursor-based DAS adaptively searches for a quantization bit for each layer. The DAS process may be accelerated via an alternative approximate optimization process, which is designed for mixed quantization scheme of a DNN model. In embodiments, a new loss function is used in the search process to simultaneously optimize accuracy and parameter size of the model. In a quantization step, the closest two integers to the cursor may be adopted as the bits to quantize the DNN together to reduce the quantization noise and avoid the local convergence problem.

Abstract: The present application provides a method, an apparatus, a device and a system for processing commodity identification and a storage medium, where the method includes: receiving image information transmitted by a camera apparatus and a distance signal transmitted by a distance sensor corresponding to the camera apparatus; determining a start frame and an end frame for a pickup behavior of a user according to the image information and the distance signal; and determining, according to the start frame and the end frame for the pickup behavior of the user, information of a commodity taken by the user. By performing a commodity identification on the start frame and the end frame for the pickup behavior of the user, and determining the information of the commodity taken by the user, commodity identification efficiency is effectively improved.

Abstract: The present application provides a method, an apparatus, a device and a system for processing commodity identification and a storage medium, where the method includes: receiving image information transmitted by a camera apparatus and a distance signal transmitted by a distance sensor corresponding to the camera apparatus; determining a start frame and an end frame for a pickup behavior of a user according to the image information and the distance signal; and determining, according to the start frame and the end frame for the pickup behavior of the user, information of a commodity taken by the user. By performing a commodity identification on the start frame and the end frame for the pickup behavior of the user, and determining the information of the commodity taken by the user, commodity identification efficiency is effectively improved.

Abstract: An information processing method, a device, a system, and a storage medium. The information processing method includes: an AI camera first obtains real-time data in a unmanned retail scenario and performs a front-end processing on the real-time data based on a neural network model, where the front-end processing includes any one or more of commodity identifying and human body monitoring, and then transmits a result of the front-end processing to a server, where the result of the front-end processing is used to trigger the server to perform face recognition and/or determine a flow direction of a commodity according to the result of the front-end processing. The cost of the entire unmanned retail distributed system and the pressure on data transmission bandwidth can be reduced, and system scalability as well as the performance of the solution to the unmanned retail can be improved effectively.