|
--- |
|
language: |
|
- en |
|
tags: |
|
- image-classification |
|
license: mit |
|
datasets: |
|
- imagenet |
|
--- |
|
|
|
|
|
# Change Detection Models for National Infrastructure Monitoring |
|
|
|
This repository contains a collection of Fine-tuned change detection models developed by Team-1 from San Jose State University as part of the National Infrastructure Monitoring project. |
|
|
|
## Models and Contributors |
|
|
|
Our team has implemented several state-of-the-art change detection models: |
|
|
|
1. **ChangeViT**: Built by Nihar |
|
- Combines Vision Transformer (ViT) and CNN architectures |
|
- Excels at detecting both large-scale and fine-grained changes |
|
- [Nihar's LinkedIn](https://www.linkedin.com/in/nihar-palem-1b955a183/) |
|
|
|
2. **BITCD**: Developed by Charishma |
|
- Uses a transformer-based approach for advanced change detection |
|
- Processes images as compact token sets for improved efficiency |
|
- [Charishma's LinkedIn](https://www.linkedin.com/in/sai-charishma-kurmala-080983128/) |
|
|
|
3. **ChangeFormer**: Implemented by Keerthana |
|
- Transformer-based architecture for satellite imagery change detection |
|
- Captures long-range spatial and temporal dependencies |
|
- [Keerthana's LinkedIn](https://www.linkedin.com/in/keerthana-raskatla-1573781a4/) |
|
|
|
4. **Multi-Modal Adaptation Network**: Content generation by Anbu |
|
- Combines optical and SAR imagery for robust change detection |
|
- Utilizes domain adaptation to align features from different image types |
|
- [Anbu's LinkedIn](https://www.linkedin.com/in/anbuvalluvan/) |
|
|
|
5. **Siamese Nested UNet**: Developed by Harika |
|
- Combines Siamese network and U-Net architectures |
|
- Excels at image comparison tasks for change detection |
|
- [Harika's LinkedIn](https://www.linkedin.com/in/harika-boyina/) |
|
|
|
## Key Features |
|
|
|
- Advanced change detection capabilities for high-resolution satellite imagery |
|
- Utilization of transformer-based approaches for capturing long-range relationships |
|
- Efficient processing of large-scale datasets |
|
- Combination of multiple imaging modalities for improved accuracy |
|
- Scalability to handle various image sizes and resolutions |
|
|
|
These models represent cutting-edge approaches in remote sensing and change detection, specifically tailored for national infrastructure monitoring applications. |
