Signboards Detection From Street-view Image Using Convolutional Neural Network: A Case Study in Vietnam
Dao Le Quang, Kien Vu Sy, Hung Luu Viet, Son Pham Bao, Hung Bui Quang · 2022 RIVF International Conference on Computing and Communication Technologies (RIVF) · 2022
In this paper, we refer to the problem of signboard detection at the instance level from street-view images. What makes this task challenging is the very heterogeneous appearance and dense distribution of signboards in urban areas. While many object detection datasets such as COCO, Pascal VOC, etc have been established, none of them pay great attention to signboard detection problems, especially in developing cities where buildings are not well planned. In this paper, a dataset of Streetview images for signboard detection is constructed covering the Cau Giay district, Hanoi, Vietnam. The dataset consists of 1217 images with signboards appearing in various shapes, sizes, and materials. Baselines are provided using a popular object detection model as Mask R-CNN with backbones Resnet-50 and Mobilenet-v2. Experimental results show that the Mask R-CNN model with Resnet-50 can achieve comparable results.