# Leveraging Deep Neural Network Compression Techniques for Real-Time
Hyperspectral Image Processing in Edge AI


Authors: Dheeraj Kumar

## Introduction

### Objective

To improve hyperspectral image analysis by integrating SSCNet \[1\] with
the FasterAI \[4\] compression technique, demonstrating efficiency and
performance on the HySpecNet-11k dataset. Hyperspectral imaging provides
rich spectral information across numerous bands, supporting applications
like remote sensing, agriculture, and medical imaging. However, the high
volume and computational demands of hyperspectral data necessitate
innovative compression and processing techniques.

### Challenges

- Managing large-scale hyperspectral datasets.
- Balancing reconstruction quality and compression efficiency.

### Contributions

- Integration of SSCNet with FasterAI pruning for compressing HSI
  compression model. to reduce model size while preserving high-quality
  image reconstruction.
- Validated on HySpecNet-11k, a large-scale hyperspectral benchmark
  dataset.
- Achieved significant reduction in model size and computational load
  with minor performance trade-offs.

## Methodology

### Dataset: HySpecNet-11k

HySpecNet-11k \[2\] is a large-scale hyperspectral dataset containing
11,483 image patches (128×128 pixels with 224 spectral bands) derived
from e Environmental Mapping and Analysis Program (EnMAP) satellite
data. It is designed for benchmarking learning-based compression and
analysis methods.

- Dataset Splits: Training (70%), Validation (20%), Test (10%).
- Preprocessing: Removed water vapor-affected bands, applied
  normalization, and used both patchwise and tilewise splits.

### Model: Spectral Signals Compressor Network (SSCNet)

SSCNet \[3\] uses 2D convolutions to compress spatial dimensions while
preserving spectral integrity.

- Encoder: 2D convolutional layers with parametric ReLU activation and
  three max-pooling.
- Decoder: Uses transposed convolutions for reconstruction.
- Compression Ratio (CR): Defined by latent channels in bottleneck
  layer.

### FasterAI Pruning Compression Technique

- Remove redundant weights or neurons.
- Fine-tune to recover performance.
- Outcome: Smaller, faster model with minimal accuracy loss.

## Experimental Results

- Peak Signal-to-Noise Ratio (PSNR): The pruned SSCNet achieved a PSNR
  of 36.09 dB, demonstrating strong reconstruction fidelity.
- Spectral Angle (SA): The pruned model maintained a low Spectral Angle
  deviation of 3.54°, indicating minimal distortion in spectral data.
- Structural Similarity Index (SSIM): The compression model achieved an
  SSIM of 0.9119 vs. the original Model SSIM is 0.9747, confirming
  preservation of spatial structures.
- Original images have 16 bpppc and compressed images have 2.53 bpppc.

### Model Efficiency

- Model Compression: The pruning process reduced the model size from
  52.59 MB to 30.20 MB (a 42.6% reduction) and the number of parameters
  from 13,783,242 to 7,911,383 (a 42.6% reduction). This highlights the
  significant decrease in computational overhead and storage
  requirements.
- VRAM Usage: The pruned model drastically reduced VRAM consumption from
  910.00 MB to 113.07 MB, making it highly efficient for deployment on
  devices with limited GPU resources.

### Comparative Analysis

- Performance Superiority: The pruned SSCNet outperformed traditional
  and learning-based compression methods in terms of computational
  efficiency and memory usage, while maintaining competitive
  reconstruction quality.
- Preservation of Fidelity: Despite pruning, the model retained
  high-quality reconstructions with:
  - PSNR ensuring strong pixel-level accuracy.
  - SSIM confirming minimal degradation in structural similarity.
  - Low spectral angle deviation, validating accurate spectral
    information preservation.

![](./images/Results.png)

## Conclusion

- Effective reduction in memory footprint and computational demands for
  real-time edge AI deployment.
- Enables practical deployment of hyperspectral models in
  resource-constrained environments.
- Supports scalable analysis for large datasets like HySpecNet-11k.

## Future Work

- Test FasterAI compression on additional hyperspectral models.
- Explore dynamic pruning strategies.
- Apply other model compression techniques.

## References

- \[1\] M. H. P. Fuchs and B. Demir, “Hyspecnet-11k: A large-scale
  hyperspectral dataset for benchmarking learning-based hyperspectral
  image compression methods,” in IGARSS 2023-2023 IEEE International
  Geoscience and Remote Sensing Symposium, IEEE, 2023, pp. 1779–1782.
- \[2\] M. H. P. Fuchs and B. Demir, “HySpecNet-11k: A large-scale
  hyperspectral benchmark dataset.” Dryad, p. 63608947808 bytes,
  Jun. 26, 2023. doi: 10.5061/DRYAD.FTTDZ08ZH.
- \[3\] R. La Grassa, C. Re, G. Cremonese, and I. Gallo, “Hyperspectral
  data compression using fully convolutional autoencoder,” Remote
  Sensing, vol. 14, no. 10, p. 2472, 2022.
- \[4\] “FasterAI,” fasterai. Available:
  https://nathanhubens.github.io/fasterai/
- \[5\] Fuchs, M. H. P., & Demir, B. (2023). HySpecNet-11k: A
  Large-Scale Hyperspectral Dataset for Benchmarking Learning-Based
  Hyperspectral Image Compression Methods. arXiv preprint
  arXiv:2306.00385v2.