Image Format Conversion for AI Training Datasets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI deep learning model training requires uniform image formats, but existing technologies lack efficient methods for converting and annotating images across different formats, such as VGG Image Annotator and Microsoft Common Objects in Context datasets, which hinders the accuracy and effectiveness of model training.
Innovation Solution
A device and system for image format conversion that includes a processor and storage, utilizing conversion rules and mapping tables to transform point coordinates from one format to another, specifically using a first and second conversion rule to handle rectangular and polygon annotation sites, enabling the conversion of images from VGG Image Annotator to Microsoft Common Objects in Context dataset format.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image format conversion is performed manually or without standardized rules, then flexibility in handling different annotation formats is maintained, but conversion accuracy and efficiency deteriorate
Solution Approach 1:
The patent transforms image annotation data by changing coordinate system parameters and data structure parameters. It converts coordinates from one format's coordinate system to another's, and transforms the organizational structure of annotation data to match different dataset requirements, thereby achieving accurate format conversion through systematic parameter transformation
Solution Approach 2:
The patent introduces a standardized conversion rule system as an intermediary between different image annotation formats. This conversion rule acts as a mediator that translates between VGG Image Annotator format and Microsoft Common Objects in Context dataset format, enabling accurate coordinate and data structure transformation without direct manual conversion
2Productivity
If multiple image formats are supported without conversion, then processing speed is maintained, but model training effectiveness deteriorates due to format inconsistency
Solution Approach 1:
The patent performs image format conversion as a preliminary action before model training. By converting all images to a unified format in advance using automated conversion rules, the system ensures format consistency is established beforehand, eliminating the need for format checking during training and maintaining high processing speed while ensuring training effectiveness
3Productivity
If automated conversion rules are implemented, then conversion efficiency is improved, but handling of complex annotation formats deteriorates
Solution Approach 1:
The patent segments the image annotation conversion process into distinct components: coordinate system transformation, data structure reorganization, and format-specific rule application. By dividing the conversion process into manageable segments, the system can efficiently handle complex annotation formats through systematic, modular conversion rules that can be applied automatically
Data Source
AI summary
A method for converting images to a unified image format for artificial intelligence deep training comprises acquiring a plurality of images and annotating the plurality of images. An annotation site of the plurality of images is determined to be quadrilateral or polygonal. The format of the annotated images is converted according to a first conversion rule when the annotation site of the annotated training images or the annotated verification images is found to be quadrilateral and the format of the annotated images is converted according to a second conversion rule when the annotation site of the annotated training images or the annotated verification images is found to be not a quadrilateral. A device employing the method is also disclosed.