Camera frame rates: blurs due to fast animal movements


My current camera records at about 12 frames per second. Due to the nature of the experiments that I’m conducing, it involves fast rodent movements. In the recorded videos, some of these movements are blurry. Is there a “blur limit” beyond which deeplabcut would have trouble picking up features?

Thank you.

Hello Aks0121,
Do you have to use a particular camera for your experiments? Even a cell phone films faster than that.
And welcome to the forum.

Thank you for your reply. That’s the current camera we have. Obviously, we don’t have unlimited resources at the lab, so I’m looking for a sweet spot when it comes to frame rate so we can get a camera that can deliver that.

Hello Aks0121 once again,
It is the camera which causes the blur as you know so just about any modern video camera should work. The standard is 30 frames per second, and many others can be adjusted over a wide range. Start with any video camera. All software used with video can adjust to the speed of acquisition but the camera is what determines the rate. Just for practice try your cell phone camera.
Any other thoughts, just ask.

Hello again,
I would look up the camera manufacturer on the ‘Web’ and try to find, or ask for a manual. Apparently you have a camera that has a frame rate adjustment or it is set in the wrong acquisition mode. The blurs are caused by how long the shutter is allowing acquisition, not on the frame rate.
It seems to be just an issue of settings, otherwise the camera could not do what is currently occurring. It would be worth checking out because you may have a very good camera now.

That’s a great point, thank you. I’ll check it out.

Hello again,
The ISO setting is what adjusts the shutter ‘speed’ if you speed it up (go to higher value) that will help quite a bit. Different camera manufactures use different terms for frame rate, I don’t know why, but you might be able to figure it out in the on screen menu if the camera has one. You can’t hurt it so experiment with the settings.


I downloaded a Mac app called Webcam Settings, and with that was able to manipulate the exposure time of the webcam. I’m happy to say that the videos are not blurry where the object moves rapidly. The only minor issue is that, whenever the object moves fast, it’s dimensions change a bit in the video (for example, when my hand was the object and I moved it fast, the recorded video shows my fingers to get thin and elongated during the rapid movements). But this is a very minor issue.

Thank you again for the suggestion.

FYI, the PS3Eye camera is dirt cheap, can run at * 640×480 @ 60 Hz or * 320×240 @ 120 Hz. It is kind of difficult, but you can get it to run on Windows, and even Windows Matlab. I haven’t yet tried Mac, but will do so soon, I guess. I’ll be publishing work to show how to use this for DLC, windows and linux (and mac, I guess), how to hold them using 3D printing and gooseneck tubing, and how to sync multiple cameras for 3D analysis. Then hopefully for real time as well.

1 Like

got it, thank you. Please keep us updated if you can.

Yeah, I agree with @ONECore - for ~9 dollars plus a lens and a 30 dollar driver for multi-cam acquisition ( for Windows it’s a nice solution!

Here is some code to run them in Python:

and a paper that uses it:


Here is instructions on 3D printing cool holders for the PS3Eye camera, how to hack it for sync’ing (kind of, I don’t think it is necessary: 0.15 ms delay between two cameras isn’t so bad), novel lens installation, and use on a variety of OS (no $30 driver required for Windows, but you’ll probably find better results with Ubuntu (over 90 FPS at 640x320 timestamped)).

There is another post on use a 3D analysis workflow (used with these cameras!). Sorry, it currently requires a step that utilizes Matlab.

Please email me with questions/problems/suggestions and acknowledge the ONE Core (email and suggested wording in the link).

1 Like