Global and Local Attention-Based Transformer for Hyperspectral Image Change Detection

Description:

We are seeking an experienced deep learning practitioner or research-oriented developer to implement a Global and Local Attention-Based Transformer model for Hyperspectral Image Change Detection (HSI-CD). This project was a research project, and now we are looking to bring it to life as a fully working prototype with all essential components.

What we need:

A fully functional codebase for the proposed model.

Implementation of both global and local attention mechanisms within a transformer architecture.

Support for standard hyperspectral image change detection datasets (e.g., LEVIR-CD, WHU-CD, etc.).

Performance metrics such as Precision, Recall, F1-Score, Kappa Coefficient, and Change Maps visualization.

A well-structured, documented, and reproducible code (preferably in PyTorch or TensorFlow).

Integration of training, validation, and testing pipelines, including data preprocessing.

Guidance on environment setup, model usage, and inference.

Any existing limitations or future enhancement suggestions are welcome.

Deliverables:

Source code (with README and documentation)

Trained model weights

Jupyter Notebook/Colab demo (optional but preferred)

Report or brief summary of the implementation and results

Ideal Candidate Should Have:

Prior experience in Transformer-based models

Experience in HSI or remote sensing image analysis

Strong understanding of attention mechanisms

Ability to deliver clean, modular, and well-documented