I’m exploring cameras for a computer vision project where I want to run AI models in real time to do sport analysis. One of the goals is to track the players position in real time. After some initial research I learned that ultra low-latency cameras with high FPS (4K, 100FPS and up) are probably a bit too expensive for this project. The price of 4K/100FPS quickly gets into the $5K+ range.
I’m looking for a camera with the following specs:
- high framerate (100FPS)
- large resolution, preferrably 4K (but I think the price tags for 4K are too high for this project)
- 2nd best 1920x1080 resolution
- price up to $1500-$2000 max
- wired connection of at least 15M
- real time streaming (h264, raw, …)
I’ve just started looking into this, so all advice is very welcome.
Thanks!
high FPS and high resolution means high data rate.
expect neural networks to take their sweet time processing all that.
visual analysis for sports has been done for decades. most of that didn’t involve AI, or only after traditional analysis reduced the data.
15 meters of cable might rule out USB, or make that a headache. expect to look for “GigE Vision” or whatever its 10GigE successor is.
are you sure 100 fps is good enough? it’s not much faster than 60 fps, and if that’s too slow, then 100 will be too slow as well.
consider the speed of things in your scene. what’s the fastest thing? how far does it travel in 1 s / 100 fps = 10 ms?
don’t forget your lighting. you’ll want strobe light synchronized (genlocked) to the camera’s exposure intervals. if done right, that could solve all motion blur issues, regardless of frame rate.