Mobile-Captured Pharmaceutical Medication Packages

1) A total of 3900 medication pack images were captured at different random angles. There are 26 images per pack. 2) Backgrounds are made of paper, cardboard, and plastic. 3) No transparent, reflexive, or patterned backgrounds are used. 4) 300 images were captured while enabling the flashlight. 5) 3600 images were captured in different lighting conditions without using a flashlight. Some images were even captured at dawn time to try the worst lighting conditions. 6) Pharmaceutical packs only were captured. Nothing was captured without its pack. 7) Captured medication packs contain tablets, capsules, syrups, creams, ampoules, gels, drops, ointments, and other types. Two types of special medical tapes were also captured. 8) 150 different pharmaceutical packages were captured. If there is a difference in the dose, size, or type, it is considered a different pack. 9) Some images were captured by different people after receiving the instructions to simulate a real patient experience. 10) Images were captured using 6 devices: • Huawei P30 Lite • Huawei Cun-L21 • Samsung A50 • Samsung A30s • iPhone XS Max • iPhone 11 Pro Max 11) Some medication packs have handwriting. 12) Some images contain shadows and flash burns. 13) 13) Some packs have different types of damage such as having spills, being ripped, etc. 14) Some images are slightly blurred. 15) All captures were taken in all day and night times for several days to try as much lighting and shadow conditions as possible. 16) The images are in the following resolutions: • 3264 x 2448 pixels – 72 dpi • 4000 x 3000 pixels – 72 dpi • 5632 x 4224 pixels – 96 dpi • 4032 x 3024 pixels – 72 dpi • 4032 x 2268 pixels – 72 dpi Images with dimensions that are greater than or equal to 4000 pixels were reduced by 20% for storing purposes.

1) 共计采集3900张药品包装图像，拍摄角度均为随机选取，每个药品包装对应26张图像。
2) 拍摄背景采用纸张、硬纸板及塑料材质。
3) 未使用透明、反光或带有花纹的背景。
4) 其中300张图像为开启闪光灯模式下采集所得。
5) 剩余3600张图像均在未开启闪光灯的不同光照条件下采集，部分图像于黎明时段拍摄，以覆盖极端光照场景。
6) 仅采集带有完整外包装的药品，未拍摄无包装的药品本体。
7) 采集的药品包装涵盖片剂、胶囊剂、糖浆剂、乳膏剂、安瓿瓶、凝胶剂、滴剂、软膏剂等多种剂型，同时包含2类特殊医用胶带包装。
8) 共计采集150种不同的药品包装：若药品的剂量、规格尺寸或剂型存在差异，则视为不同的包装品类。
9) 部分图像由不同人员在接收拍摄规范后完成采集，以模拟真实患者的使用场景。
10) 图像采集设备共计6款，分别为：华为P30 Lite、华为Cun-L21、三星A50、三星A30s、iPhone XS Max、iPhone 11 Pro Max。
11) 部分药品包装带有手写标识。
12) 部分图像存在阴影及闪光灯过曝痕迹。
13) 部分包装存在不同类型的损坏，如洒漏、撕裂等。
14) 部分图像存在轻微模糊现象。
15) 采集工作于多日的昼夜时段内开展，以覆盖尽可能多的光照与阴影场景。
16) 图像分辨率涵盖以下规格：• 3264×2448像素，72dpi • 4000×3000像素，72dpi • 5632×4224像素，96dpi • 4032×3024像素，72dpi • 4032×2268像素，72dpi。为便于存储，所有长边≥4000像素的图像均按20%比例压缩。

搜集汇总

背景与挑战

背景概述

该数据集是一个用于计算机视觉研究的药物包装图像集合，包含3900张图像，覆盖150种不同的药物包装类型，采集自多种手机设备，并在多样化条件下拍摄，如不同光照、角度和背景，以模拟真实世界场景。其特点在于包含了手写文字、阴影、包装损坏等现实挑战，适用于药物识别和分类模型的训练与评估。