18284076. Neural Network Inference Acceleration Method, Target Detection Method, Device, and Storage Medium simplified abstract (BOE TECHNOLOGY GROUP CO., LTD.)

From WikiPatents
Jump to navigation Jump to search

Neural Network Inference Acceleration Method, Target Detection Method, Device, and Storage Medium

Organization Name

BOE TECHNOLOGY GROUP CO., LTD.

Inventor(s)

Chunshan Zu of Beijing (CN)

Neural Network Inference Acceleration Method, Target Detection Method, Device, and Storage Medium - A simplified explanation of the abstract

This abstract first appeared for US patent application 18284076 titled 'Neural Network Inference Acceleration Method, Target Detection Method, Device, and Storage Medium

Simplified Explanation

The patent application describes a method for accelerating neural network inference by automatically performing model compression, graph optimization, and deployment optimization on a neural network model using an accelerated data set to obtain an accelerated neural network model.

  • Acquiring a neural network model and an accelerated data set
  • Automatically performing model compression, graph optimization, and deployment optimization on the neural network model
  • Model compression techniques include model quantification, model pruning, and model distillation
  • Graph optimization involves optimizing the directed graph of the neural network model
  • Deployment optimization focuses on optimizing the deployment platform of the neural network model
  • Performing inference evaluation on the accelerated neural network model

Potential Applications

This technology can be applied in various fields such as image recognition, natural language processing, autonomous vehicles, and healthcare for faster and more efficient neural network inference.

Problems Solved

1. Slow inference speed of neural networks 2. High computational resources required for neural network inference

Benefits

1. Improved inference speed 2. Reduced computational resources usage 3. Enhanced efficiency of neural network models

Potential Commercial Applications

Optimized neural network inference acceleration technology can be utilized in industries such as e-commerce, finance, cybersecurity, and manufacturing for real-time data processing and decision-making.

Possible Prior Art

Prior art in neural network acceleration includes techniques such as hardware accelerators, parallel processing, and software optimizations to improve the speed and efficiency of neural network inference.

What are the potential limitations of this technology?

One potential limitation of this technology could be the complexity of implementing the various optimization techniques on different neural network models, which may require specialized knowledge and expertise.

How does this technology compare to existing solutions?

This technology offers a comprehensive approach to accelerating neural network inference by combining model compression, graph optimization, and deployment optimization techniques, which can result in significant improvements in speed and efficiency compared to traditional methods that focus on individual optimizations.


Original Abstract Submitted

A neural network inference acceleration method includes: acquiring a neural network model to be accelerated and an accelerated data set; automatically performing accelerating process on the neural network model to be accelerated by using the accelerated data set to obtain the accelerated neural network model, wherein the accelerating process includes at least one of the following: model compression, graph optimization and deployment optimization, wherein the model compression includes at least one of the following: model quantification, model pruning and model distillation, wherein the graph optimization is the optimization for the directed graph of the neural network model to be accelerated, and the deployment optimization is the optimization for the deployment platform of the neural network model to be accelerated; and performing inference evaluation on the accelerated neural network model.