Related Experiment Video
Updated: Jun 30, 2026
08:58
Artificial Intelligence Approaches to Assessing Primary Cilia
Published on: May 1, 2021
3.5K
Artificial Intelligence Approach for Classifying Images of Upper-Atmospheric Transient Luminous Events.
1Department of Electrical and Computer Engineering, University of Puerto Rico at Mayaguez, Mayagüez, PR 00681, USA.
Sensors (Basel, Switzerland)
|May 25, 2024
Summary
Automating the classification of Transient Luminous Events (TLEs) using deep learning models like CNNs and Vision Transformers significantly improves efficiency. ResNet50 demonstrated the highest accuracy in identifying these upper-atmospheric phenomena.
Area of Science:
- Atmospheric physics
- Geophysics
- Computer science
Background:
- Transient Luminous Events (TLEs) are brief optical phenomena in the upper atmosphere linked to thunderstorms.
- Manual classification of TLEs is challenging due to their rapid and unpredictable nature.
Purpose of the Study:
- To develop and evaluate automated methods for classifying TLEs.
- To compare the performance of Convolutional Neural Networks (CNNs) and Vision Transformers (ViT) for TLE classification.
Main Methods:
- Utilized state-of-the-art deep learning models: Vision Transformer (ViT), ResNet50, ResNet18, GoogLeNet, and SqueezeNet.
- Employed data augmentation (rotation, translation, flipping) and image preprocessing (bilateral filtering).
- Evaluated models based on classification accuracy and execution time.
Main Results:
- All tested models achieved high classification accuracy for TLEs.
- ResNet50 exhibited the highest accuracy among the evaluated architectures.
- A trade-off between classification accuracy and execution speed was observed across models.
Conclusions:
- Deep learning, particularly transfer learning with pre-trained CNNs, is effective for automated TLE classification.
- Model selection requires balancing accuracy and computational efficiency based on specific application needs.
