If you take a picture of the pill with the camera, it is an application that informs you about the pill information.
We started development as a capstone design project at Korea Transportation University, and we are currently developing server replacement and update.

working structure
Application (Client) ↔ Main Server ↔ AI Server

1. Taking pictures of pills in the application
2. Send photos to the main server after touching the search button
3. Photo transfer from main server to artificial intelligence server
4. Extract pill information after analyzing photos in artificial intelligence server
5. Send the extracted pill information to the main server
6. Search by entering pill information on the website of the Pharmaceutical Information Service on the main server
7. Crawl the pharmaceutical information source search results and send them to the application


AI model used
ㆍText-Detection
: Naver Clova-ai CRAFT pretrained model

ㆍText-recognition
: Naver Clova-ai deep-text-recognition-benchmark
: Pretrained model(TRBA-case-sensitive)+fine-tuning with custom_data(croped pill text image)

ㆍShape-classification
: Efficient-Net-Pytorch