Application of novel convolutional neural network-based artificial intelligence in counting system
Autor: LIN, YU-CHEN LIN, HSUAN-YI LIN, YU-JOU WU, YING-HAN HSIAO, WEI-WEN
Company: National Taiwan University National Tsing Hua University National Taiwan University of Science and Technology
Country: Taiwan
e-mail: eddie@wiipa.org.tw
web: https://www.wiipa.org.tw/
We develop a new generation of artificial intelligence bacteria and cell technology platform, and use convolutional neural network to identify bacteria and cells, and combine big data with deep learning to improve the accuracy of identification.
This convolutional neural network can effectively achieve low-error and real-time colony calculation, and it can be used with the USB camera module (IPEVO DO-CAM) to take pictures and measure at any time. By analyzing the image and adding the database, the system can judge the number of a picture of thousands of colonies within one second, and the error is within 10.
For the repeatability and accuracy of the experiment, our high-accuracy and low-cost artificial intelligence cell counting system, taking Escherichia coli as an example, residual learning combined with the overall error value of the convolutional neural network of the global pooling layer Within ±2 bacteria.
In the future, the system will be extended to detect gastric cancer in endoscopic images, estimate the impact of human papillomavirus type on the risk of recurrence of cervical dysplasia, classify skin cancer, identify microbial volatile organic compound signatures, and detect fractures. High accuracy but low cost applied to cancer cell identification, benefiting the general public.
