Browsing by Author "Salim, Md. Shaeak Ibna"

Now showing 1 - 2 of 2

Single and multi-view 2D image processing for enhanced 3D object reconstruction
(BRAC University, 2024-05) Ghosh, Swapnil; Mahmud, Md. Muhtasim; Ahmed, Asrar; Turjo, Tashdid Al Shafi; Salim, Md. Shaeak Ibna; Alam, Md. Ashraful
This paper presents an efficient approach for 3D object reconstruction using Single and multi- view 2D image processing. In real-world scenarios, it also focuses on practical applications. Our approach is about the advanced image processing techniques as well as deep learning models which convert multiple 2D views of an object into a detailed 3D model. Our method is a novel application of convolutional neural networks that merge features from each view for ensuring consistent geometry and texture in the final model. Additionally, we introduce a robust merging module based on CNN. It improves the model’s fidelity by focusing on areas with significant detail variation across different views. Our tests on lots of challenging datasets show that our method enhances computational efficiency as well as it has significant potential for practical applications in areas such as virtual reality, augmented reality, and automated quality control in manufacturing. This research marks a significant step forward in digital imaging and computer vision, offering new possibilities for industry and technology advancements.
Single and multi-view 2D image processing for enhanced 3D object reconstruction
(BRAC University, 2024-05) Ghosh, Swapnil; Mahmud, Md. Muhtasim; Ahmed, Asrar; Turjo, Tashdid Al Shafi; Salim, Md. Shaeak Ibna; Alam, Md. Ashraful
This paper presents an efficient approach for 3D object reconstruction using Single and multi- view 2D image processing. In real-world scenarios, it also focuses on practical applications. Our approach is about the advanced image processing techniques as well as deep learning models which convert multiple 2D views of an object into a detailed 3D model. Our method is a novel application of convolutional neural networks that merge features from each view for ensuring consistent geometry and texture in the final model. Additionally, we introduce a robust merging module based on CNN. It improves the model’s fidelity by focusing on areas with significant detail variation across different views. Our tests on lots of challenging datasets show that our method enhances computational efficiency as well as it has significant potential for practical applications in areas such as virtual reality, augmented reality, and automated quality control in manufacturing. This research marks a significant step forward in digital imaging and computer vision, offering new possibilities for industry and technology advancements.