Fall Rescuers

• 

  sample detection image by the prediction server

• 

• 

• 

We aim to address the phenomenon of slow medical response towards accidental falls among elderly who live alone in Singapore by making use of volunteers from NGO organizations. Therefore, our theme chosen is Giving Back. Our hack also considers the first question of Giving back where elderly are able to receive faster care when injured and volunteers are also able to volunteer anywhere and anytime based on their location. We have trained a YOLOV5 model with fallen and non fallen human datasets and made use of augmentations on these datasets such as gaussian blur and perspective distortion to capture elderly falls from all angles and distances. Cameras can be set up in the elderly person's home, taking snapshots of the room every 5 minutes. The image would then be sent to a server to check if the person is fallen or standing. If the elderly is detected as fallen, his details and photo would be uploaded onto the webapp. Volunteers can check the list of locations that require help. If the elderly is sleeping and the detection was a false alarm, volunteers can also mark the alert as a false alarm. Should the volunteer choose to help the elderly, a keylock code to a locker below the elderly's house containing a key to the elderly's house would be provided. This provides elderly people with immediate care and also allows volunteers to help out anytime anywhere.

NLP emotion detection was also done but was not added into the final webapp

Built With

• fastapi
• flask
• html
• python
• torch
• yolov5