Publication: Universal barcode localization and decoding using YOLOV7 algorithm
| datacite.subject.fos | oecd::Engineering and technology::Electrical engineering, Electronic engineering, Information engineering::Electrical and electronic engineering | |
| dc.contributor.author | Stephen, Wong Wei Feng | |
| dc.date.accessioned | 2025-05-13T04:11:34Z | |
| dc.date.available | 2025-05-13T04:11:34Z | |
| dc.date.issued | 2023-08 | |
| dc.description.abstract | The implementation of barcode technology has revolutionized manufacturing processes by enhancing supply chain management, boosting productivity, and minimizing errors. Barcode localization and decoding represents a cutting-edge skill, aiming to develop a rapid and precise method of barcode reading. Dynamsoft Barcode Reader (DBR) is a barcode reader equipped algorithms that deliver unmatched speed, accuracy, and read rates in barcode decoding. However, challenging conditions such as excessive light exposure, low contrast, or poor illumination in the industrial environment can pose difficulties for DBR. To address this, deep learning techniques can be employed for barcode detection, thereby improving detection accuracy. Instead of analysing the entire image, which may contain multiple barcodes, a deep learning model can be trained to detect and crop individual barcodes into a single image, which can then be fed into DBR Python API for decoding. This project trained three prominent deep learning algorithms, namely You Only Look Once (YOLOv5), YOLOv7, and improved YOLOv7, for barcode detection. Through the project, it is found that all YOLO models had achieved an average precision and recall of 99.5% and 99.8% respectively. For inference speed, YOLOv7 tiny has achieved a remarkable speed of 556 FPS with an inference time of 1.8ms. This concluded that YOLOv7 tiny has the highest speed compared to YOLOv5 (278 FPS) and YOLOv7 (256 FPS). For DBR decoding, two methods are used, which are DBR Online Demo SDK and DBR Python API and achieved a decoding accuracy of 100% and 98.9% for Set 1 and Set 2 images respectively. As a conclusion, the implementation of YOLOv7 tiny with DBR Python API has resulted in a significant improvement. Thus, the thesis has successfully met its objectives. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/21596 | |
| dc.language.iso | en | |
| dc.title | Universal barcode localization and decoding using YOLOV7 algorithm | |
| dc.type | Resource Types::text::report::technical report | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |