---
license: mit
task_categories:
- image-to-image
- other
tags:
- hyperspectral-imaging
- hsi-compression
- image-compression
- image-reconstruction
- deep-learning
- wireless-communications
pretty_name: BTC-Net V2 Dataset
size_categories:
- 1K<n<10K
---

# BTC-Net V2: Spatial-Priority Hierarchical Reconstruction for Hyperspectral Image Compression

This repository contains the officially released dataset for **BTC-Net V2**. 

For the complete implementation, network architectures, and pre-trained models, please visit our official GitHub repository:
👉 **[GitHub: MWan-deeplearner/BTC-Net-V2](https://github.com/MWan-deeplearner/BTC-Net-V2)**

---

## 📌 Dataset Overview

This dataset consists of processed Hyperspectral Image (HSI) data (specifically from the AVIRIS sensor), structured and chunked to facilitate deep learning-based image compression and hierarchical wireless transmission research.

* **Data Format:** MATLAB `.mat` files packed into sequential `.zip` chunks.
* **Total Samples:** 1,195 processed HSI patches/sub-volumes.
* **Core Application:** Spatial-priority hierarchical reconstruction, HSI compression, deep learning.

---

## 📂 Repository Structure & Naming Convention

To prevent download interruptions caused by network fluctuations, the 1,195 `.mat` files are distributed evenly into **12 independent zip archives** (approx. 100 files per chunk).

The repository is structured as follows:

```text
.
├── README.md                           # This Dataset Card
└── aviris_mat_chunk_01.zip            # Sub-dataset chunk 01 (Contains files 1-100)
├── aviris_mat_chunk_02.zip            # Sub-dataset chunk 02 (Contains files 101-200)
...
└── aviris_mat_chunk_12.zip            # Sub-dataset chunk 12 (Contains remaining files)