Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained Likelihood
With the exploration of outer space, the number of space targets has increased dramatically, while the pressures of space situational awareness have also increased. Among them, spacecraft recognition is the foundation and a critical step in space situational awareness. However, unlike natural images...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-09-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/15/17/4321 |
_version_ | 1797581923242999808 |
---|---|
author | Xi Yang Dechen Kong Ren Lin Dong Yang |
author_facet | Xi Yang Dechen Kong Ren Lin Dong Yang |
author_sort | Xi Yang |
collection | DOAJ |
description | With the exploration of outer space, the number of space targets has increased dramatically, while the pressures of space situational awareness have also increased. Among them, spacecraft recognition is the foundation and a critical step in space situational awareness. However, unlike natural images that can be easily captured using low-cost devices, space targets can suffer from motion blurring, overexposure, and excessive dragging at the time of capture, which greatly affects the quality of the images and reduces the number of effective images. To this end, specialized or sufficiently versatile techniques are required, with dataset diversity playing a key role in enabling algorithms to categorize previously unseen spacecraft and perform multiple tasks. In this paper, we propose a joint dataset formulation to increase diversity. Our approach involves reformulating two local processes to condition the Conditional Neural Adaptive Processes, which results in global feature resampling schemes to adapt a pre-trained embedding function to be task-specific. Specifically, we employ variational resampling to category-wise auxiliary features, adding a generative constraint to amortize task-specific parameters. We also develop a neural process variational inference to encode representation, using grid density for conditioning. Our evaluation of the BUAA dataset shows promising results, with no-training performance close to a specifically designed learner and an accuracy rate of 98.2% on unseen categories during the joint training session. Further experiments on the Meta-dataset benchmark demonstrate at least a 4.6% out-of-distribution improvement compared to the baseline conditional models. Both dataset evaluations indicate the effectiveness of exploiting dataset diversity in few-shot feature adaptation. Our proposal offers a versatile solution for tasks across domains. |
first_indexed | 2024-03-10T23:13:33Z |
format | Article |
id | doaj.art-9397a5fe5aa94b169cd94a84ad86d33c |
institution | Directory Open Access Journal |
issn | 2072-4292 |
language | English |
last_indexed | 2024-03-10T23:13:33Z |
publishDate | 2023-09-01 |
publisher | MDPI AG |
record_format | Article |
series | Remote Sensing |
spelling | doaj.art-9397a5fe5aa94b169cd94a84ad86d33c2023-11-19T08:47:36ZengMDPI AGRemote Sensing2072-42922023-09-011517432110.3390/rs15174321Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained LikelihoodXi Yang0Dechen Kong1Ren Lin2Dong Yang3State Key Laboratory of Integrated Services Networks, School of Telecommunications Engineering, Xidian University, Xi’an 710071, ChinaState Key Laboratory of Integrated Services Networks, School of Telecommunications Engineering, Xidian University, Xi’an 710071, ChinaState Key Laboratory of Integrated Services Networks, School of Telecommunications Engineering, Xidian University, Xi’an 710071, ChinaXi’an Institute of Space Radio Technology, Xi’an 710100, ChinaWith the exploration of outer space, the number of space targets has increased dramatically, while the pressures of space situational awareness have also increased. Among them, spacecraft recognition is the foundation and a critical step in space situational awareness. However, unlike natural images that can be easily captured using low-cost devices, space targets can suffer from motion blurring, overexposure, and excessive dragging at the time of capture, which greatly affects the quality of the images and reduces the number of effective images. To this end, specialized or sufficiently versatile techniques are required, with dataset diversity playing a key role in enabling algorithms to categorize previously unseen spacecraft and perform multiple tasks. In this paper, we propose a joint dataset formulation to increase diversity. Our approach involves reformulating two local processes to condition the Conditional Neural Adaptive Processes, which results in global feature resampling schemes to adapt a pre-trained embedding function to be task-specific. Specifically, we employ variational resampling to category-wise auxiliary features, adding a generative constraint to amortize task-specific parameters. We also develop a neural process variational inference to encode representation, using grid density for conditioning. Our evaluation of the BUAA dataset shows promising results, with no-training performance close to a specifically designed learner and an accuracy rate of 98.2% on unseen categories during the joint training session. Further experiments on the Meta-dataset benchmark demonstrate at least a 4.6% out-of-distribution improvement compared to the baseline conditional models. Both dataset evaluations indicate the effectiveness of exploiting dataset diversity in few-shot feature adaptation. Our proposal offers a versatile solution for tasks across domains.https://www.mdpi.com/2072-4292/15/17/4321spacecraft recognitionfew-shot feature adaptationgenerative familyneural processes |
spellingShingle | Xi Yang Dechen Kong Ren Lin Dong Yang Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained Likelihood Remote Sensing spacecraft recognition few-shot feature adaptation generative family neural processes |
title | Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained Likelihood |
title_full | Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained Likelihood |
title_fullStr | Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained Likelihood |
title_full_unstemmed | Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained Likelihood |
title_short | Generalizing Spacecraft Recognition via Diversifying Few-Shot Datasets in a Joint Trained Likelihood |
title_sort | generalizing spacecraft recognition via diversifying few shot datasets in a joint trained likelihood |
topic | spacecraft recognition few-shot feature adaptation generative family neural processes |
url | https://www.mdpi.com/2072-4292/15/17/4321 |
work_keys_str_mv | AT xiyang generalizingspacecraftrecognitionviadiversifyingfewshotdatasetsinajointtrainedlikelihood AT dechenkong generalizingspacecraftrecognitionviadiversifyingfewshotdatasetsinajointtrainedlikelihood AT renlin generalizingspacecraftrecognitionviadiversifyingfewshotdatasetsinajointtrainedlikelihood AT dongyang generalizingspacecraftrecognitionviadiversifyingfewshotdatasetsinajointtrainedlikelihood |