Synthetic Dataset for Object-to-Model Deep Learning

本研究创建了一个名为Synthetic Dataset for Object-to-Model Deep Learning的数据集，由帝国理工学院计算系开发。该数据集包含100,000张合成图像，用于训练深度卷积神经网络（CNN）以识别超市产品。数据集通过3D模型和渲染技术生成，使用摄影测量技术从真实世界对象获取3D模型。此数据集旨在解决工业应用中特定对象识别数据稀缺的问题，通过自动像素标注简化了数据标注过程，适用于实时对象定位和分类任务。

This study creates a dataset named "Synthetic Dataset for Object-to-Model Deep Learning", developed by the Department of Computing, Imperial College London. This dataset comprises 100,000 synthetic images intended for training deep convolutional neural networks (CNNs) to recognize supermarket products. The dataset is generated using 3D models and rendering techniques, where the 3D models are acquired from real-world objects via photogrammetry. This dataset aims to address the issue of data scarcity for specific object recognition in industrial applications, simplifies the data annotation process through automatic pixel-level labeling, and is applicable to real-time object localization and classification tasks.