000			03940nas a2200421 a 4500
005			20260817161526.0
008			260112s20249999th u ms t 000 eng d
035			_a.b12469816
099		9	_aAIT Diss. no.ISE-24-02
100	1		_aWuttinan Sereethavekul
245	1	0	_aAdaptive lightweight license plate image recovery using deep learning based on generative adversarial network
260			_aPathum Thani, Thailand : _bAsian Institute of Technology, _c2024
300			_a109 leaves : _bill. +1 online resource
490	1		_aDissertation ; _vno. ISE-24-02
500			_aA dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Microelectronics and Embedded Systems
502			_aThesis (Ph. D.) - Asian Institute of Technology, 2024
520			_aMany Convolutional Neural Networks (CNNs) methods have already surpassed tradi tional approaches to image restoration tasks. Those CNNs models were usually de signed to enhance single tasks such as an image resolution (super-resolution) or image denoising, but we came up with unconventional goals, that is, multiple recovery tasks from a single network design. Although the Transformer design has recently gained at tention in image recovery tasks, they are too slow. In order to work with license plate images from a traffic camera stream, the system has to be responsive. So, we proposed a fast and lightweight deep learning-based data recovery system using a Generative Ad versarial Network (GAN) principle named License Plate Recovery GAN (LPRGAN). The design has a proposed encoder-decoder style inspired by an autoencoder aided by dual classification networks. This style suits problem-characteristic learning because strong contextual information is retrieved from the down-scaled representations. This proposed system has three main features such as identifying a problem, data recovery, and fail-safe mechanism. The core of system is a data recovery unit (LPRGAN), is used to recover license plate images from multiple degraded input images. Most existing im age restoration systems do not have self-awareness, leading to an inefficiency problem. Unlike existing works, this system has anomaly detection and will only process on a de graded input, reducing workload overhead, improving efficiency and a fail-safe feature that prevents an unexpected bad output. Hence, the proposed algorithm requires less resource to deploy on a low-power machine such as edge computing devices, opening up newpossibilities in on-device computing. Our proposed research can recover several degraded problems up to 720p resolution at 15 frames per second on a single graphic card, 256x128 resolution at 17 frames per second on a CPU-only workstation machine, or 7 frames per second on an ultra-low-power tablet PC.
650		0	_aMachine Learning
650		0	_aNeural networks (Computer science)
650		0	_aData recovery (Computer science)
700	0		_aMongkol Ekpanyapong, _eChairperson
700	1		_aDailey, Matthew N., _eExamination Committee
700	0		_aHuynh, Trung Luong, _eExamination Committee
710	2		_aRoyal Thai Government, _eScholarship Donor
710	2		_aAIT Fellowship, _eScholarship Donor
810	2		_aAsian Institute of Technology. _tDissertation ; _vno. ISE-24-02
856	4	0	_3Full-Text _uhttp://203.159.5.9/ait-thesis/detail.php?q=B23315
907			_a.b12469816 _bmnarc _ca
902			_a260122
998			_b0 _c260116 _dm _eh _fa _g0
945			_lmnarc
945			_lmnarc
942			_c40
942			_c61
909			_aBarcode : 30050120422661 _bCREATED : 2026-12-01 _cRECORD # : i13565473 _dLPATRON : 0 _eLCHKIN : - _f# RENEWALS : 0 _g# OVERDUE : 0 _hIUSE3 : 0 _iTOT CHKOUT : 0 _jTOT RENEW : 0
909			_aBarcode : - _bCREATED : 2026-12-01 _cRECORD # : i13565485 _dLPATRON : 0 _eLCHKIN : - _f# RENEWALS : 0 _g# OVERDUE : 0 _hIUSE3 : 0 _iTOT CHKOUT : 0 _jTOT RENEW : 0
999			_c1610 _d1610