Medeval
MedEval uses a machine learning model to audit real-time hospital footage, verifying the presence of crucial amenities & providing instant audits to ensure that hospitals meet standard requirements
Created on 28th April 2024
•
Medeval
MedEval uses a machine learning model to audit real-time hospital footage, verifying the presence of crucial amenities & providing instant audits to ensure that hospitals meet standard requirements
The problem Medeval solves
MedEval simplifies the task of monitoring hospital facilities by leveraging machine learning to analyze real-time footage. It enables conducting random checks, ensuring hospitals maintain all necessary equipment at all times and also eliminates the need for physical presence. This approach reduces the likelihood of malpractices during audits, promoting transparency and accountability in healthcare facilities.
Challenges we ran into
Finding a relevant dataset: We found a dataset (Hospital Indoor Object Detection) depicting hospital room images which consisted of approximately 5000 images from a GitHub profile. The dataset had a lot of redundant images and few images that weren’t relevant for our problem statement.
Data Cleaning: The dataset had to be cleaned and reduced to 2641 images, only including images of hospital rooms and equipment. This made the dataset more efficient and easy to train.
Model Integration: We faced difficulties in integrating in the YOLOv5 model to our web application. YOLOv5 trains on weights and does not provide the necessary architecture for integration.
Technologies used