Ensemble convolutional neural network for classifying holograms of deformable objects.

Area of Science:

• Computer Vision
• Machine Learning
• Holography

Background:

• Ensemble deep learning invariant hologram classification (EDL-IHC) has been demonstrated for classifying holograms of deformable objects.
• Existing EDL-IHC methods require significant computational resources for high success rates (≥99%).
• There is a need for low-complexity deep learning networks (DLNs) that achieve high success rates in hologram classification.

Purpose of the Study:

• To propose a low-complexity deep learning network (DLN) for hologram classification.
• To improve the success rate of hologram classification compared to existing methods.
• To reduce the computational resources required for accurate hologram classification.

Main Methods:

• Development of a novel, low-complexity deep learning network (DLN) for hologram classification.
• The proposed network is also referred to as ensemble deep learning invariant hologram classification (EDL-IHC).
• Evaluation of the proposed EDL-IHC method on the classification of holograms of handwritten numerals.

Main Results:

• The proposed low-complexity EDL-IHC method achieved a 2.86% increase in success rate for classifying holograms of handwritten numerals.
• The new EDL-IHC approach demonstrates improved performance over previous methods.
• The developed DLN offers a more computationally efficient solution for hologram classification.

Conclusions:

• The proposed low-complexity EDL-IHC method offers a significant improvement in hologram classification accuracy.
• This approach provides a computationally efficient alternative for classifying holograms of deformable objects.
• The findings suggest potential for wider adoption of DLNs in holographic data analysis.