Datasets
Standard Dataset
Chinese Industrial Parts Multimodal Dataset
- Citation Author(s):
- Submitted by:
- Ke Wang
- Last updated:
- Wed, 12/11/2024 - 05:13
- DOI:
- 10.21227/3ngv-br47
- Data Format:
- License:
- Categories:
- Keywords:
Abstract
This dataset comprises images of parts from real industrial scenarios and virtual reality environments. Real images are sourced from actual industrial settings, ensuring both authenticity and diversity, while virtual reality images, which make up approximately 11% of the dataset, are captured through precise 3D modeling. Approximately 30% of the part information was manually authored by industry experts, while the remaining 70% was generated by multimodal large models such as Wenxin Yiyan and GPT-4. Our dataset is 5.47GB in size, and you can obtain it from this Baidu Netdisk link: https://pan.baidu.com/s/1FGmOpw8mANiCzmhxrXURTA?pwd=fgj6
This dataset spans across 6 distinct categories, with each category comprising 1000 images along with several close-up videos of parts. Each line in the text file corresponds to a description of an image, totaling 1000 lines of descriptions that align with the order of the images.