Summary of Heuristical Comparison Of Vision Transformers Against Convolutional Neural Networks For Semantic Segmentation on Remote Sensing Imagery, by Ashim Dahal et al.
Heuristical Comparison of Vision Transformers Against Convolutional Neural Networks for Semantic Segmentation on Remote Sensing Imagery
by Ashim Dahal, Saydul Akbar Murad, Nick Rahimi
First submitted to arxiv on: 14 Nov 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper explores the application of Vision Transformers (ViT) for semantic segmentation in remote sensing aerial images. ViT has shown great promise in computer vision tasks, particularly in image classification and segmentation. The study compares three key factors: using a weighted fused loss function to optimize performance, transfer learning with Meta’s MaskFormer versus generic UNet Convolutional Neural Network (CNN), and the trade-offs between these models and current state-of-the-art segmentation models. The results show that the novel combined weighted loss function significantly improves CNN model performance compared to transfer learning with ViT. This research has implications for the use of ViT in semantic segmentation tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well a new type of AI model, called Vision Transformers (ViT), can do when it comes to separating objects from backgrounds in aerial images taken from space or planes. These models have been very good at recognizing things like animals and cars in pictures. The study compares different ways of using these models to see what works best. The results show that one way of using the model is better than another way, which can help make better decisions when working with these kinds of images. |
Keywords
» Artificial intelligence » Cnn » Image classification » Loss function » Neural network » Semantic segmentation » Transfer learning » Unet » Vit
