Fast And Easy Sensor Adaptation With Selftraining

• doi: 10.1109/ACCESS.2023.3238875
• title: Fast and Easy Sensor Adaptation With Self-Training
• publisher: IEEE
• isbn:
• issn: 2169-3536
• rank: 2482
• access_type: CCBY - IEEE is not the copyright holder of this material. Please follow the instructions via https://creativecommons.org/licenses/by/4.0/ to obtain full-text articles and stipulations in the API documentation.
• content_type: Journals
• abstract: Object detectors based on deep neural networks have the disadvantage that new labels should be acquired whenever the complementary metal-oxide semiconductor (CMOS) image sensor (CIS) is changed. In this study, we propose a fast and easy two-step sensor-adaptation method without labels for the target domain; 1) simple adaptation, and 2) self-training. The simple-adaptation process transfers the knowledge of the source model to the target model by updating the batch normalization parameters, and matches the feature distributions of the source domain and those of target domain. In the self-training process, we employ the ensemble model strategy to mitigate the over-fitting problem using noisy pseudo labels generated by the simple-adaptation model. Quantitative and qualitative experiments show that the proposed method can transfer the knowledge from one CIS model to another, even if the data format of the target domain is different from that of the source CIS domain.
• article_number: 10024297
• pdf_url: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10024297
• html_url: https://ieeexplore.ieee.org/document/10024297/
• abstract_url: https://ieeexplore.ieee.org/document/10024297/
• publication_title: IEEE Access
• conference_location:
• conference_dates:
• publication_number: 6287639
• is_number: 10005208
• publication_year: 2023
• publication_date: 2023
• start_page: 8870
• end_page: 8877
• citing_paper_count: 0
• citing_patent_count: 0
• download_count: 434
• insert_date: 20230123
• index_terms:
  • ieee_terms:
    • Semi-supervised learning
    • Adaptation models
    • Data models
    • Noise measurement
    • Detectors
    • Solid modeling
    • Semiconductor device modeling
  • author_terms:
    • Domain adaptation
    • raw image
    • self-training
    • semi-supervised learning
    • sensor adaptation
  • dynamic_index_terms:
    • Neural Network
    • Deep Neural Network
    • Data Format
    • Object Detection
    • Cognitive Model
    • Psychological Models
    • Knowledge Model
    • Simplified Version
    • Ensemble Model
    • Image Sensor
    • Camera Sensor
    • Target Domain
    • Quantitative Experiments
    • Source Model
    • Source Domain
    • Pseudo Labels
    • Noisy Labels
    • False Positive
    • False-positive
    • Training Set
    • False Negative
    • False-negative
    • Validation Set
    • Raw Files
    • Domain Adaptation
    • Raw Images
    • Student Network
    • Student Model
    • Detection Performance
    • RGB Images
    • Average Precision
    • Faster R-CNN
    • Target Dataset
    • Target Data Set
    • Upper Threshold
• authors:
  • Author Name: Jinhyuk Choi
    Affiliation: Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea
    Author URL: https://ieeexplore.ieee.org/author/37089715599
    ID: 37089715599
    Order: 1
    Author Affiliations:
    • Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea
  • Author Name: Byeongju Lee
    Affiliation: Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea
    Author URL: https://ieeexplore.ieee.org/author/37089711268
    ID: 37089711268
    Order: 2
    Author Affiliations:
    • Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea
  • Author Name: Seho Shin
    Affiliation: Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea
    Author URL: https://ieeexplore.ieee.org/author/37089715738
    ID: 37089715738
    Order: 3
    Author Affiliations:
    • Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea
  • Author Name: Daehyun Ji
    Affiliation: Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea
    Author URL: https://ieeexplore.ieee.org/author/37089708446
    ID: 37089708446
    Order: 4
    Author Affiliations:
    • Samsung Advanced Institute of Technology, Yeongtong-gu, Suwon-si, Gyeonggi-do, South Korea

Image Sensor

• sensor_type: CMOS
• resolution: 16-bit raw
• dynamic_range: Not specified in the text
• pixel_size: Not specified in the text
• dark_current: Not specified in the text

Optical Data

• focal_length: Not specified in the text
• aperture: Not specified in the text
• field_of_view: Not specified in the text
• distortion: Not specified in the text

Performance Metrics

• frame_rate: Not specified in the text
• signal_to_noise_ratio: Not specified in the text
• sensitivity: Not specified in the text
• shutter_speed: Not specified in the text
• power_consumption: Not specified in the text
• noise: Not specified in the text

Applications & Benefits

• cell_imaging: Not specified in the text
• benefits: Facilitates digital imaging in smartphones, medical devices, and automotive systems.

Supporting Organizations

• supported_by: Samsung Advanced Institute of Technology, Gyeonggi-do, South Korea

Manuscript Details

• publication_date: 2023

Relevancy Score

• score: 8
• missing_fields:
  • fill factor
  • quantum efficiency
  • readout speed
  • temporal noise
  • fixed-pattern noise
  • analog-to-digital conversion techniques
  • microlenses
  • on-chip colour filters
  • global or rolling shutters
  • backside illumination

Processed JSON Filename

request_fc711a7e-f9a4-4d22-b5b4-298a0de147f0-fast_and_easy_sensor_adaptation_with_selftraining.json