Neural Network Imaging Device Layered Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional imaging devices face increased processing time and memory usage due to the execution of image recognition functions, leading to inefficiencies in memory allocation and latency.
Innovation Solution
An imaging device is designed with a sensor that outputs first image data, a first processing unit that processes a first layer of a neural network on smaller units of image data, and a second processing unit that processes a second layer on the output from the first processing unit, allowing for efficient decomposition of processing tasks in units of lines rather than entire frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image recognition function is executed on entire first image data, then recognition accuracy is improved, but processing time increases and memory region increases
Solution Approach 1:
The patent divides the image recognition processing into multiple layers (first layer, second layer, etc.) where each layer processes the image data independently. The first processing unit executes processing of the first layer on the first image data, and the second processing unit executes processing of the second layer on the output from the first processing unit. This segmentation allows parallel processing and reduces the time each unit needs to process, thereby reducing overall processing time while maintaining recognition accuracy through multi-layer analysis.
Solution Approach 2:
The patent introduces a dimensional transformation by converting spatial image data into feature space through multiple processing layers. Each layer transforms the input data into a different representation, extracting features at different levels of abstraction. This dimensional transformation enables the system to maintain high recognition accuracy by analyzing data in multiple feature dimensions rather than processing the entire spatial image data directly.
2Measurement precision
If image recognition function is executed on entire first image data, then recognition accuracy is improved, but available memory region decreases
Solution Approach 1:
The patent segments the image data processing into multiple independent layers, where each layer processes a portion of the recognition task. The first processing unit handles the first layer processing on the input image data, and the second processing unit handles the second layer processing on the intermediate results. This segmentation reduces the peak memory requirements compared to processing the entire image data in a single unit, as each processing unit only needs to hold the data relevant to its specific layer.
Solution Approach 2:
The patent extracts intermediate features from the image data through the first processing unit before passing them to the second processing unit. Instead of loading and processing the entire original image data through all processing steps, the system extracts essential features at each layer and passes only these extracted features to the next layer, thereby reducing the memory region required for processing.
3Reliability
If processing is executed in units of entire first image data, then processing completeness is improved, but processing time increases
Solution Approach 1:
The patent divides the processing task into multiple segments (layers) that can be executed in parallel by different processing units. The first processing unit executes processing of the first layer on the first image data, while the second processing unit executes processing of the second layer on the output from the first processing unit. This segmentation enables concurrent processing of different layers, improving processing speed while maintaining completeness through the systematic progression through all necessary processing layers.
Solution Approach 2:
The patent implements continuous processing where the output of one layer becomes the input of the next layer without interruption. The first processing unit continuously processes the first layer on incoming image data, and the second processing unit continuously processes the second layer on the output from the first processing unit. This continuous flow of data through multiple processing stages maintains processing completeness while improving overall processing speed through pipelined execution.
Data Source
AI summary
A memory region is suppressed. An imaging device according to the present disclosure includes: a sensor that outputs first image data; a first processing unit that executes processing of a first layer in a neural network having a layered structure on the first image data in units of second image data having a size smaller than an entire size of the first image data; and a second processing unit that executes processing of a second layer in the neural network on a processing result output from the first processing unit.