BdSLImset (Bangladeshi Sign Language Image Dataset)

孟加拉手语图像数据集（BdSLImset）是一个包含不同孟加拉手语图像的数据集。孟加拉手语（BdSL）是孟加拉听力受损人群常用的交流媒介。开发一个实时系统来检测图像中的这些迹象是一项巨大的挑战。在本文中，我们提出了一种从实时执行的图像中检测 BdSL 的技术。我们的方法使用基于卷积神经网络的对象检测技术来检测图像区域中是否存在符号并识别其类别。为此，我们采用了 Faster Region-based Convolutional Network 方法并开发了一个数据集 BdSLImset 来训练我们的系统。以前检测 BdSL 的研究工作通常依赖于外部设备，而大多数其他基于视觉的技术不能实时有效地执行。然而，我们的方法没有这些限制，实验结果表明，所提出的方法成功地实时识别和识别孟加拉国标志。

The Bangladesh Sign Language Image Dataset (BdSLImset) is a dataset containing images of various Bangladesh Sign Language (BdSL) gestures. Bangladesh Sign Language (BdSL) is a commonly used communication medium for hearing-impaired populations in Bangladesh. Developing a real-time system to detect these gestures in images poses significant challenges. In this paper, we propose a technique for detecting BdSL from real-time captured images. Our method uses convolutional neural network-based object detection techniques to detect the presence of signs in image regions and recognize their categories. To this end, we adopt the Faster Region-based Convolutional Network approach and develop the BdSLImset dataset to train our system. Previous research efforts for BdSL detection typically rely on external devices, while most other vision-based techniques cannot perform efficiently in real time. However, our method avoids these limitations. Experimental results demonstrate that the proposed method can successfully recognize and classify Bangladesh sign gestures in real time.

搜集汇总

背景与挑战

背景概述

BdSLImset是一个孟加拉手语图像数据集，专为实时检测孟加拉手语而设计。该数据集采用Faster R-CNN等卷积神经网络技术进行对象检测，以训练系统实现高效识别。它由阿萨努拉科技大学于2018年发布，并附有相关论文和开源代码。