LabelScr: Automated Image Annotation on Mobile and Web Applications for Object Detection and Localization Model Pipelines
Chandan Jha, Gourav Kumar Shrivastava · 2022
Object localization and detection, as an emerging and demanding subject in the computer vision community, is critical for creating next-generation computer vision systems, and over the past decade, it has received a lot of attention. Deep learning requires large amounts of data, and building datasets for this task is not a simple operation. Recent developments in the field of computer vision are mostly driven by the use of deep learning technologies. Detailed processes with several phases and users with varied roles are common in larger benchmark datasets during annotation. However, in smaller projects with limited resources, this can be difficult to implement. Therefore, this work presents a process for creating an image dataset for mobile apps (Android/iOS) and web apps without the need for manual annotation of images. These datasets can be used to build models that can detect and localize icons and other objects available on digital screens. Segmentation of different parts on digital screens can also be achieved using a dataset created using this method.