CHORD

该数据集是一个包含室内场景平面图和布局的大型数据集，由专业室内设计师创建了9,706个设计方案。该数据集涵盖了家居物品和房间配置的全面内容，无需进行数据清洗或预处理。它提供了26个超级类别的家居物品，包括家具、固定装置和电器。其规模达到9,706个场景，大约是3D-FRONT数据集的1.4倍。该数据集的任务是用于3D室内场景合成和布局生成。This dataset is a large-scale collection containing indoor scene floor plans and layouts, with 9,706 design schemes developed by professional interior designers. It covers comprehensive content of household items and room configurations, eliminating the need for data cleaning or preprocessing. The dataset provides 26 super-categories of household items, including furniture, fixtures and electrical appliances. With a total of 9,706 scenes, its scale is approximately 1.4 times that of the 3D-FRONT dataset. The tasks supported by this dataset are 3D indoor scene synthesis and layout generation.

搜集汇总

背景与挑战

背景概述

CHORD（CHOrus Recognition Dataset）是一个用于多模态副歌识别的数据集，旨在提升歌曲搜索的准确性。该数据集与ICANN2021论文相关，支持基于PyTorch和BERT-wwm-ext中文预训练模型的训练与测试，适用于音乐信息检索任务。