Improving underwater object classification: BC-ViT

Kurz, Aidan; Adams, Ethan; Depoian, Arthur C.; Kim, Hae Jin; Bailey, Colleen P.; Guturu, Parthasarathy

doi:10.1117/12.2619134

Cited by 7 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent research on automatic target detection has transitioned to focus on machine learning based approaches using modern techniques such as advanced sensors, [6][7][8][9] deep learning, [10][11][12] convolutional neural networks, 2, 13, 14 and more recently vision transformers. 15 There are two sensors commonly used to capture information for ATD/R systems: Synthetic Aperture Radars (SAR) and Infrared (IR). 2,9 These sensors collect data, typically in the form of a frame-by-frame video in various environments allowing a multitude of targets to be seen then fed into neural networks to accomplish tasks of detection, recognition, and motion prediction.…”

Section: Introductionmentioning

confidence: 99%

“…This paper proposes an efficient target detection network that utilizes the overarching theme of the SegFormer most notably the MIX FFN and multi scale resolution as well as the convolution based transformer block introduced in BC-ViT. 15 The merger of these two networks allow for tiny network, Edge IR Vision Transformer (EIR-ViT), at approximately 115 thousand parameters that is tasked with the sole purpose to detect cars utilizing the FLIR dataset for training and validation.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automatic target detection utilizing an edge IR vision transformer (EIR-ViT)

Adams

Depoian

Kurz

et al. 2023

Automatic Target Recognition XXXIII

View full text Add to dashboard Cite

The detection and recognition of targets within imagery and video analysis is vital for military and commercial applications. The development of infrared sensor devices for tactical aviation systems imagery has increased the performance of target detection. Due to the advancements of infrared sensors capabilities, their use for field operations such as visual operations (visops) or reconnaissance missions that take place in a variety of operational environments have become paramount. Many techniques implemented stretch back to 1970, but were limited due to computational power. The AI industry has recently been able to bridge the gap between traditional signal processing tools and machine learning. Current state of the art target detection and recognition algorithms are too bloated to be applied for on ground or aerial mission reconnaissance. Therefore, this paper proposes Edge IR Vision Transformer (EIR-ViT), a novel algorithm for automatic target detection utilizing infrared images that is lightweight and operates on the edge for easier deployability.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Automatic target detection utilizing an edge IR vision transformer (EIR-ViT)

Adams

Depoian

Kurz

et al. 2023

Automatic Target Recognition XXXIII

View full text Add to dashboard Cite

show abstract

“…This paper proposes a new architecture, MHATT, for 1D classification of time-series data. Insights were drawn from techniques employed by BC-ViT, 7 which advanced the field of 2D classification, for the creation of this architecture. MHATT uses highly modified vision transformers to improve performance while keeping the network highly efficient.…”

Section: Introductionmentioning

confidence: 99%

Augmented multi-head classification network: MHATT

Cayce

Depoian

Bailey

et al. 2023

Signal Processing, Sensor/Information Fusion, and Target Recognition XXXII

View full text Add to dashboard Cite

Classification of one-dimensional (1D) data is important for a variety of complex problems. From the finance industry to audio processing to the medical field, there are many industries that utilize 1D data. Machine learning techniques have excelled at solving these classification problems, but there is still room for improvement because the techniques have not been perfected. This paper proposes a novel architecture called Multi-Head Augmented Temporal Transformer (MHATT) for 1D classification of time-series data. Highly modified vision transformers were used to improve performance while keeping the network exceptionally efficient. To showcase its efficacy, the network is applied to heartbeat classification using the MIT-BIH OSCAR dataset. This dataset was ethically-split to ensure a fair and intensive test for networks. The novel architecture is 94.6% more efficient and had a peak accuracy of 91.79%, which was a 13.6% reduction in error over a recent state-of-the-art network. The impressive performance and efficiency of the MHATT architecture can be exploited by edge devices for unmatched performance and flexibility of deployment.

show abstract

“…To handle poor lighting and weather, EIR-ViT utilizes thermal images that are not affected by these conditions. To handle low resolution, we leverage the BC-ViT [12], which was able to obtain remarkable binary classification scores with a massively decimated input image.…”

Section: Introductionmentioning

confidence: 99%

Neural Network Classifiers for Object Detection in Optical and Infrared Images

Adams

View full text Add to dashboard Cite

This thesis presents a series of neural network classifiers for object detection in both optical and infrared images. The focus of this work is on efficient and accurate solutions. The thesis discusses the evolution of the highly efficient and tiny network Binary Classification Vision Transformer (BC-ViT) and how through thoughtful modifications and improvements, the BC-ViT can be utilized for tasks of increasing complexity. Chapter 2 discusses the creation of BC-ViT and its initial use case for underwater image classification of optical images. The BC-ViT is able to complete its task with an accuracy of 99.29\% while being comprised of a mere 15,981 total trainable parameters. Chapter 3, Waste Multi-Class Vision Transformer (WMC-ViT), introduces the usefulness of mindful algorithm design for the realm of multi-class classification on a mutually exclusive dataset. WMC-ViT shows that the task oriented design strategy allowed for a network to achieve an accuracy score of 94.27\% on a five class problem while still maintaining a tiny parameter count of 35,492. The final chapter demonstrates that by utilizing functional blocks of BC-ViT, a simple and effective target detection algorithm for infrared images can be created. The Edge Infrared Vision Transformer (EIR-ViT) showed admirable results with a high IoU of 79\% coupled with a massive reduction in parameters compared to current state of the art solutions at roughly 120,000. This thesis ends with summary of Chapters 2, 3, and 4 that further discusses the implications this design strategy offers.

show abstract

Improving underwater object classification: BC-ViT

Cited by 7 publications

References 0 publications

Automatic target detection utilizing an edge IR vision transformer (EIR-ViT)

Automatic target detection utilizing an edge IR vision transformer (EIR-ViT)

Augmented multi-head classification network: MHATT

Neural Network Classifiers for Object Detection in Optical and Infrared Images

Contact Info

Product

Resources

About