-
Transferability of Convolutional Neural Networks in Stationary Learning Tasks
Authors:
Damian Owerko,
Charilaos I. Kanatsoulis,
Jennifer Bondarchuk,
Donald J. Bucci Jr,
Alejandro Ribeiro
Abstract:
Recent advances in hardware and big data acquisition have accelerated the development of deep learning techniques. For an extended period of time, increasing the model complexity has led to performance improvements for various tasks. However, this trend is becoming unsustainable and there is a need for alternative, computationally lighter methods. In this paper, we introduce a novel framework for…
▽ More
Recent advances in hardware and big data acquisition have accelerated the development of deep learning techniques. For an extended period of time, increasing the model complexity has led to performance improvements for various tasks. However, this trend is becoming unsustainable and there is a need for alternative, computationally lighter methods. In this paper, we introduce a novel framework for efficient training of convolutional neural networks (CNNs) for large-scale spatial problems. To accomplish this we investigate the properties of CNNs for tasks where the underlying signals are stationary. We show that a CNN trained on small windows of such signals achieves a nearly performance on much larger windows without retraining. This claim is supported by our theoretical analysis, which provides a bound on the performance degradation. Additionally, we conduct thorough experimental analysis on two tasks: multi-target tracking and mobile infrastructure on demand. Our results show that the CNN is able to tackle problems with many hundreds of agents after being trained with fewer than ten. Thus, CNN architectures provide solutions to these problems at previously computationally intractable scales.
△ Less
Submitted 21 July, 2023;
originally announced July 2023.
-
Deterministic Multi-sensor Measurement-adaptive Birth using Labeled Random Finite Sets
Authors:
Jennifer Bondarchuk,
Anthony Trezza,
Donald J. Bucci Jr
Abstract:
Measurement-adaptive track initiation remains a critical design requirement of many practical multi-target tracking systems. For labeled random finite sets multi-object filters, prior work has been established to construct a labeled multi-object birth density using measurements from multiple sensors. A truncation procedure has also been provided that leverages a stochastic Gibbs sampler to truncat…
▽ More
Measurement-adaptive track initiation remains a critical design requirement of many practical multi-target tracking systems. For labeled random finite sets multi-object filters, prior work has been established to construct a labeled multi-object birth density using measurements from multiple sensors. A truncation procedure has also been provided that leverages a stochastic Gibbs sampler to truncate the birth density for scalability. In this work, we introduce a deterministic herded Gibbs sampling truncation solution for efficient multi-sensor adaptive track initialization. Removing the stochastic behavior of the track initialization procedure without impacting average tracking performance enables a more robust tracking solution more suitable for safety-critical applications. Simulation results for linear sensing scenarios are provided to verify performance.
△ Less
Submitted 12 July, 2023;
originally announced July 2023.
-
Multi-Target Tracking with Transferable Convolutional Neural Networks
Authors:
Damian Owerko,
Charilaos I. Kanatsoulis,
Jennifer Bondarchuk,
Donald J. Bucci Jr,
Alejandro Ribeiro
Abstract:
Multi-target tracking (MTT) is a classical signal processing task, where the goal is to estimate the states of an unknown number of moving targets from noisy sensor measurements. In this paper, we revisit MTT from a deep learning perspective and propose a convolutional neural network (CNN) architecture to tackle it. We represent the target states and sensor measurements as images and recast the pr…
▽ More
Multi-target tracking (MTT) is a classical signal processing task, where the goal is to estimate the states of an unknown number of moving targets from noisy sensor measurements. In this paper, we revisit MTT from a deep learning perspective and propose a convolutional neural network (CNN) architecture to tackle it. We represent the target states and sensor measurements as images and recast the problem as an image-to-image prediction task. Then we train a fully convolutional model at small tracking areas and transfer it to much larger areas with numerous targets and sensors. This transfer learning approach enables MTT at a large scale and is also theoretically supported by our novel analysis that bounds the generalization error. In practice, the proposed transferable CNN architecture outperforms random finite set filters on the MTT task with 10 targets and transfers without re-training to a larger MTT task with 250 targets with a 29% performance improvement.
△ Less
Submitted 25 July, 2023; v1 submitted 27 October, 2022;
originally announced October 2022.
