Drug image classification with deep learning by using Fast Region-based Convolution Neural Network

• Published in: 2024 Joint International Conference on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering (ECTI DAMT & NCON) pp. 551 - 555
• Main Authors: Boonthep, Narasak, Chaiwongsai, Jirabhorn, Srisungsittisunti, Bowonsak, Udomsripaiboon, Thana
• Format: Conference Proceeding
• Language: English
• Published: IEEE 31.01.2024
• Subjects: Convolution; convolutional neural network (CNN); Deep learning; Drug identification; Drugs; Medical services; Neural networks; Packaging; Shape
• ISSN: 2768-4644
• DOI: 10.1109/ECTIDAMTNCON60518.2024.10479992

At present, there are many drugs produced for the treatment of various diseases. The most often registered drug formulations, according to statistics on drug formulation registration, were pills. Some pills have remarkably similar outward appearances, despite not being the same type, as may be observed. This makes it challenging for those without medical knowledge to categorize pills as drugs of any kind. Deep learning is a machine learning approach that is being applied in a variety of applications, including image classification, image analysis, and object recognition. In the context of identifying drug images, computer vision algorithms can be trained to recognize various characteristics such as color, shape, markings, and packaging of different drugs. This can be particularly useful in various domains such as healthcare, law enforcement, and pharmaceuticals. This paper proposes a drug image classification by using Fast Region-base convolution neural network (Fast R-CNN). Convolution layers, pooling layers, and fully connected layers are some of the building blocks that CNN employs to automatically and adaptively learn spatial hierarchies of input through backpropagation. A region-based strategy is used in this paper to reduce the detection time. Experiments have shown that our proposed framework can accurately identify different types of 20 drugs with over 98 percent accuracy.