2026-05-10T15:36:51Z https://meral.edu.mm/oai

oai:meral.edu.mm:recid/00007926 2021-12-13T02:51:55Z 1582963674932 1582963674932:1597396989070 user-miit

Transparent Object Detection Using Faster R-CNN May Phyo Khaing Ei Khaing Win "Recently, object detection has become a popular area in computer vision and object recognition. In many robotic researches, the most basic step is to perform object detection so that the reaction can be taken after detecting object location and its category. One of the main tasks for domestic robots is household object detection. In this paper, we intend to detect transparent objects such as glass in images. Compared with other kinds of objects, the detection of transparent object is very difficult to be performed using classical computer vision algorithms. Most of the classical computer vision algorithms implement the object detection based on their appearance such as colour or texture of the objects. However, the appearance of transparent objects changes according to different backgrounds and illumination conditions. With the popularity of object detection researches, deep learning algorithms now offer a high performance in detection of objects. Therefore, we apply one of the deep learning models called Faster R-CNN (Regions with Convolutional Neural Network) to perform detection of transparent objects and evaluate the performance of the system. According to experimental results, the system achieves 89.8% mAP in the detection of transparent objects. Keywords – Computer vision and object recognition, Deep learning, Domestic robots, Faster R-CNN, Transparent object detection " 2018-10-06 http://hdl.handle.net/20.500.12678/0000007926 https://meral.edu.mm/records/7926