Data and Algorithms for Detecting public perceptions of different types of urban spaces with computer vision techniques and social media big data
收藏NIAID Data Ecosystem2026-05-02 收录
下载链接：
https://data.mendeley.com/datasets/dh94v24vx5
下载链接
链接失效反馈官方服务：
资源简介：
The datasets and algorithms for Detecting public perceptions of different types of urban spaces with computer vision techniques and social media big data.
Datasets are social media images collected from public Weibo posts generated from 1 January to 31 December 2023 with the hashtags "Chengdu Global Centre" and "Chengdu Taikoo Li" (the names of two popular shopping malls in Chengdu City, Sichuan Province, China). The datasets include two parts:
1. Image-based Labelling Dataset: 9,691 images, labelled with Image-based Labelling Strategy, used for model training/evaluating/testing.
2. Research-based Labelling Dataset: 9,691 images, labelled with Research-based Labelling Strategy, used for model training/evaluating/testing.
Algorithms include:
1. Zeroshot CLIP Model: implemented on both the Image-based Labelling Dataset and the Research-based Labelling Dataset.
2. Zeroshot LLaVA Model: implemented on both the Image-based Labelling Dataset and the Research-based Labelling Dataset.
3. Fine-tuned ResNet Model: implemented on both the Image-based Labelling Dataset and the Research-based Labelling Dataset.
创建时间：
2024-09-20
