The best camera resolution and framerate for AI Failure Detection

A properly configured camera is essential in order for the AI model to detect failures. There are 3 main factors contirbuting to a well-configured camera:

Frame Rate for 3D Printer Camera

The camera framerate (FPS) does not need to be high in order to run the AI. The AI software analyzes the current image every 10 seconds. A frame rate of 1 frame per second is sufficient to run the AI software, however it is recomended to use a frame rate of 5 frames per second to run the AI software. Keep in mind that this will be the framerate shown in OctoPrint/Mainsail/DuetWebControl.

Warning A high frame rate is CPU intensive and may bog down your system, especially when configured with a high resolution and quality.

Resolution for 3D Printer Camera

The Camera should produce a good quality image of the print bed area. It is recomended to use a RPi-based camera like the Raspberry Pi Camera Module 3 or the Raspbeerry Pi Camera v2. The camera must run with a minimum resoltuion of 640 x 640 pixels. It is highly recomended to run the camera quality at a higher resolution than this, because the camera may be used for other purposes in addition to AI monitoring.

Keep in mind that the image is resized to 640 x 640 pixels, so extremely high resolution images will not make a difference.

As a result, it is recomended that the camera be run in a resolution of 1920 x 1080 pixels.

Quality

The camera streamer has a quality setting that ranges from 0-100, with 100 being the best quality. It is recomended to run the camera in a quality of 100 or as close to 100 as possible, with the absolute minimum being a quality of 80.