Human condition analysis systems

Human condition analysis systems:

Postural assessment problems are relatively difficult and important tasks at the same time, and computer vision engineers struggle to adopt methods and techniques that make the implementation of posture assessment tasks more efficient than before.

Here are some projects I've built around this field.

Fall detection:

Monitoring the condition of patients and people inside hospitals and agencies.

The fall of people, whether due to an accident or their exposure to a state of fainting, is one of the things that need a very quick intervention in order to provide first aid for these people.

As surveillance cameras become ubiquitous, it is possible to delegate some algorithms to analyze camera streams to find falls and issue appropriate alerts.

In this project, we used the methods mentioned below to analyze the video stream, where the idea behind the quality of detection is to train a custom neural network on the output of human movement analysis during specific time periods of about 30 frames,

This ensures that the machine produces intermittent reports of human movements and then understands them.

Data for walking, standing, sitting, falling, etc. was used. and tested on general data.

Methods:

RetinaNet Or Yolo -> #Person detection

OpenPose -> #Human body estimation *It is applied to Boxes discovered by Retna Net to reduce time*

Graph-CNN -> #human keyPoints understanding

Tools:

Python, C++.

Torch, Opencv, etc.