T/SHLY T/SHLY1005-2024 The guide for building a sample database of image data for artificial intelligence identification of forest management inspection

Scope

Main Technical Content: During the development process, this annotation standard mainly references documents such as "Guide for Standardization Work - Part 1: Structure and Drafting Rules for Standardization Documents", GB/T 41867 "Information Technology - Artificial Intelligence - Terminology", GB/T 17798 "Geospatial Data Sample Exchange Format", and LY/T 2930 "Standard Specifications for Forestry Data Sample Collection". The main indicators are explained as follows: 1. **Construction Content**: It mainly proposes the framework for the construction of an image data sample library, which includes three parts: defining requirements, establishing a data sample library, and managing the data sample library. - **Defining Requirements**: This includes three parts: data sample classification, data sample collection, and data sample annotation, with their standard principles specified separately. - **Data Sample Classification**: This is primarily determined by the actual needs of regulatory authorities, taking into account the basic principles of computer vision algorithms and the results of 5 rounds of algorithm iteration. The data samples are sorted according to classification principles and categorized based on specific work scenarios. - **Data Sample Collection**: This summarizes the collection experience in different seasons, regions, and methods, including pre-collection preparation, data sample types, collection methods, metadata recording, and data sample formats. - **Data Sample Annotation**: This mainly follows the principles of deep learning algorithms and actual model training and iteration experience, standardizing the basic process and quality control methods for data annotation. It also summarizes the roles, tools, objects, quality, and precautions involved to ensure data annotation quality and improve model training accuracy. - **Establishing a Data Sample Library**: This includes establishing an image data sample library, a labe

Development Information

Related Standards