Multitask Text Recognition in Images for Accurate Map Data Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image detection and recognition technologies are resource-intensive and vary in accuracy, making it costly and time-consuming to verify the effectiveness of object detection and recognition tasks, which affects the performance of image processing applications.
Innovation Solution
A machine-learned model with task-specific heads is used to detect and recognize text segments in images, determining attributes such as business names, numbers, and websites, which are then used to update map data, leveraging a multitask model with a main encoder and task-specific heads to generate attribute data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection and recognition techniques are used, then detection accuracy can be improved, but computing resources and time consumption increase significantly
Solution Approach 1:
The machine-learned model is divided into a shared backbone and multiple task-specific heads, where each head is specialized for a particular attribute detection task. This segmentation allows the model to process different attributes in parallel while sharing common feature extraction resources, thereby improving both accuracy and processing efficiency
Solution Approach 2:
The shared backbone of the machine-learned model serves multiple detection tasks simultaneously, making it a universal feature extractor that can be applied to various attribute detection problems. This multi-functionality reduces redundant computation while maintaining high detection accuracy across different tasks
2Reliability
If multiple detection tasks are performed separately, then each task can be optimized, but overall processing time and resource usage increase
Solution Approach 1:
Multiple detection tasks are merged into a single machine-learned model that processes all tasks simultaneously. The model combines multiple task-specific heads under a shared backbone, allowing parallel processing of different attributes (e.g., business names, phone numbers, websites) without requiring separate verification passes, thus reducing overall processing time while maintaining task optimization
3Measurement precision
If high-accuracy object detection is implemented, then detection reliability improves, but computing hardware requirements and costs increase
Solution Approach 1:
The model architecture parameters are optimized to achieve high detection reliability without requiring excessive computational resources. By adjusting the depth and width of the backbone network, the number and type of task-specific heads, and other architectural parameters, the system achieves accurate multi-attribute detection while controlling hardware requirements and deployment costs
Data Source
AI summary
Methods, systems, devices, and non-transitory computer readable media for processing images and updating map data are provided. The disclosed technology can include receiving image data comprising a plurality of images. A plurality of attributes associated with the plurality of images can be determined based on inputting the image data into a machine-learned model that is configured to recognize one or more text segments detected in the plurality of images. The machine-learned model can comprise a plurality of task-specific heads configured to determine the plurality of attributes. One or more entities associated with the plurality of attributes can be determined. Furthermore, attribute data comprising the plurality of attributes associated with the one or more entities can be generated. Furthermore, based on the attribute data, map data associated with a plurality of locations can be updated.