Automatic License Plate Recognition (ALPR) systems have been widely used for traffic management, law enforcement, and security. Previous ALPR methods have achieved impressive performance on single-type license plates. However, multi-type license plate recognition is still challenging due to various character layouts and fonts. This study presents an efficient ALPR system based on the YOLO-V8 object detector and optical character recognition (OCR) to read the license plate (LP) content. OCR sometimes misinterprets certain English letters as Arabic numbers and certain Arabic numbers as English letters because of their visual similarities. Therefore, this study has improved the OCR to improve recognition results by including the checking character index (CCI) stage to fine-tune the detection results. The proposed ALPR uses YOLO-v8 and was evaluated using real Iraqi license plate datasets and achieved a precision and recall of $100 \%$, and $\mathbf{9 9 . 8 \%}$ respectively, with mAP confidence curve values of 99.5.
Signal processing techniques are widely used in almost every humanitarian technology based real life applications targeted to meet the needs of a particular community. Among different HT applications, in this talk, AI-driven (machine learning (ML) and deep learning (DL) based) automatic disease detection techniques will be presented. It is found that the use of signal processing techniques on a given 1D or 2D data, prior to ML based classification, is very helpful. We have demonstrated a very effective use of acoustic characteristics of heart sounds in the ML based automatic classification of heart-valve disorders. Moreover, some of our recently proposed attention-based deep neural network architectures will be presented where we propose temporal, channel, spatial, and guided attention mechanisms for achieving better discriminative features. The results on disease detection performance will be demonstrated in dealing with cardiac (PCG, ECG) and different medical imaging data (e.g. CT, MRI, endoscopy, ultrasound).
Computer vision is a fascinating area for the AI research community. Coupled with hardware and/or other AI applications like speech interactions, the possible topic areas become even wider. An open platform will be introduced with functionalities demonstration. Further, to foster AI research collaborations, the session will touch briefly on the RoboCup@Home Education initiative.
This speech will highlight the potential of machine learning (ML) in medical image analysis and disease diagnostic. First, this speech will focus on the implementation of machine learning for disease diagnostic systems, which involves image processing and intelligent classification phases. Contrast enhancement and noise filtering are used to improve the quality of medical images, before features extraction and analysis are performed to extract significant features. Intelligent classifiers are then used for disease classification purposes. Three case studies namely, classification of cervical cancer (2 cases) and detection of sperm motility will be presented. At the end of the speech, challenges of ML based disease diagnostic system will be discussed as well. Examples of challenges are (i) gaining confidence among medical experts, (ii) competition from deep learning approaches, (iii) developing a more objective system etc.
Humanitarian technology encompasses a wide range of areas aimed at addressing humanitarian challenges to meet the needs of a particular community. The main objective of this workshop is to illustrate the key factors behind designing HT projects utilizing signal processing techniques. Unlike conventional project design, the HT projects need to address various challenges of the local community. Hence, it is very important to properly understand the needs, maximize community engagement with proper stakeholder mapping. Some major signal processing applications in humanitarian technology will be explained.
