制造资源检测图像数据集
收藏资源简介:
数据集内容描述:车间制造车间的现场照片和标记结果。对照片上的生产制造资源(人员、设备和物料)进行人工识别框选,并标记每个对象的所属类别。将框选结果和标记结果,以TXT文件的形式进行独立记录和保存。 数据来源:试验数据。 采集方式:利用加工场所现场的四部摄像头,采集监控图片,并进行数据整理;之后对整理后的图片进行人工标注,得到类别等信息;随后将所有数据统一进行存储。 加工方式:图像收集后,由数据收集人员依据使用目的,对图片先进行裁剪。然后人工识别出图片中所包含的生产制造资源,记录每一种生产制造资源的类型,并测出其在图片中的位置。将上述信息,记录于图片对应的TXT文件中,进行保存。为扩大数据集容量,对图片进行了翻转、旋转等处理,并同步调整TXT中的位置信息。数据加工进行过程中,对于所有科学数据定期进行质量控制核查。后期进行审核验收,统一审核汇总,形成数据集。 设备情况:海康威视(DS-2CD2T46FWDA4-LS),个人计算机(dell燃7000,宏碁(Acer)暗影骑士3)。 数据量:4.8GB
Dataset Content Description: On-site photographs from the workshop manufacturing floor and their corresponding annotation results. Manually identified the production and manufacturing resources (personnel, equipment and materials) in the photographs, drew bounding boxes around them, and labeled the category each object belongs to. The bounding box results and annotation results were independently recorded and saved in TXT file format. Data Source: Experimental data. Collection Method: Four cameras installed on the processing site were used to collect surveillance images, followed by data organization. Manual annotation was then performed on the organized images to obtain category information and other relevant details. Finally, all collected data was uniformly stored. Processing Method: After image collection, data collectors first cropped the images according to the intended use. Subsequently, the production and manufacturing resources contained in the images were manually identified, the type of each production and manufacturing resource was recorded, and its position within the image was measured. The aforementioned information was recorded in the TXT file corresponding to each image and saved. To expand the dataset size, image augmentation operations such as flipping and rotation were performed on the images, and the position information in the corresponding TXT files was adjusted synchronously. During the data processing phase, regular quality control inspections were conducted on all scientific data. Final review and acceptance were carried out, and all data was uniformly reviewed and consolidated to form the final dataset. Equipment Specifications: Hikvision (DS-2CD2T46FWDA4-LS), personal computers (Dell Ran 7000, Acer Nitro 3). Data Volume: 4.8 GB




