This pilot study focuses on developing an automated defect detection system for power line inspections. Traditional methods, including aerial and manual inspections, require significant human effort and are prone to inconsistencies. Our goal is to assess the feasibility of machine learning models for defect recognition, improving efficiency and reliability.
Project Scope
The study aims to analyze existing inspection data, develop a neural network-based recognition model, and evaluate its effectiveness. The project follows a structured approach divided into three phases:
-
Data Analysis: Reviewing current defect data, defining annotation standards, and assessing dataset quality.
-
Model Training: Training machine learning algorithms to classify and detect defects.
-
Evaluation & Optimization: Testing model performance, identifying limitations, and refining the approach.
|
|
The project is being developed for ČEPS, a.s.
Contact person: Adam Novozámský
