Multiple people tracking Opencv tracking of multiple people example videos and some ideas about the business behind. I would like to mention that this video is not related to my current business. This is just love for computer vision.
This short blog post should be about something else.
The problem of computer vision business potential
- People point to any mistake in the algorithm. Mistakes are visible. People don't care if the big data gives you huge statistics errors. They are not able to count the million line of something.
- They are able to count what they really see on video.
Where is the point? Even the error is visible and sometimes huge. This gives you better results than one man counting the people from the window all day long.
Are you able to track and monitor what every on screen doing count targets and write the results somewhere.
No
Do not sell the video. Sell the data
This is important. Maybe the most important thing. You are provided data about the day's activity about the year's activity.
Not the person for one day counting the traffic on the Prague town center crossing.
Video delivery for Internet of Things
You have to solve the problem with stream video delivery to the cloud as cheaply as it possible.
There is one thing. Internet of things. Maybe is a better old approach just to manage the computer vision where the camera is located.. Computers are smaller and faster. The video delivery is the problem. You have to deal with so many problems. Instead of transferring just the data.
The benefit is that the people in front of the monitor just see your mistakes.
Cost-effective computer vision
There is maybe a question of what is effective. Processing the video in the cloud has also many advantages. Scales. Updates. Your algorithm is under control. You can handle whatever resolution. This is the only issue of the cost of your solution. It is not necessary with some cameras to manage your own HW in that solution.
I have a lot of ideas like that. I spent lots of time not only on the algorithm itself. For example, another huge topic is
Self-configuration of computer vision algorithm
Yes, this one. Do not spend 3 hours or months making a perfect YouTube video. There is no reason.
The algorithm should be somehow self-configuring. And it may not be perfect. Let the rest of the world accept errors in computer vision.
Have a nice evening and let me know. What do you think?