Lithology Identification of Improved Model based on DeepLabv3 + Deep Convolutional Neural Network

Aiming at the complexity of outcrop lithology identification, an improved outcrop lithology identification algorithm based on DeepLabv3 + is proposed. We use the Xception network replaced by the MobileNetV2 module to enhance the ability to extract lithologic features and make the model more lightwei...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2024 International Conference on New Trends in Computational Intelligence (NTCI) s. 533 - 537
Hlavní autoři: Wang, Ju, Yin, Senlin, Hu, Zhangming, Bai, Kai
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 18.10.2024
Témata:
Accuracy
Attention mechanism
Attention mechanisms
Classification algorithms
Complexity theory
Computational intelligence
Computational modeling
Convolutional neural networks
Feature extraction
lightweight model
lithology identification
Semantic segmentation
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Aiming at the complexity of outcrop lithology identification, an improved outcrop lithology identification algorithm based on DeepLabv3 + is proposed. We use the Xception network replaced by the MobileNetV2 module to enhance the ability to extract lithologic features and make the model more lightweight. In order to segment pixels more accurately, the CA (Coordinate attention) attention mechanism is introduced, so that the improved backbone network can extract lithologic features more accurately in complex environments. The experimental results show that compared with the traditional DeepLabv3 +, the accuracy of the improved algorithm is increased by 1.08 percentage points, the MP A is increased by 2.4 percentage points, and the average classification accuracy is 94.86 %. Compared with other mainstream semantic segmentation models, all aspects of performance are more prominent.
DOI:10.1109/NTCI64025.2024.10776114