Hello,
I’d like to get some advice on running a depth model on the Raspberry Pi 5 with the Hailo-8L AI Hat. I already have a Raspberry Pi Camera Module v3. Besides an additional camera module, I’m wondering if I need any other hardware or software to run it. Does Hailo software support this setup? Or would it be better to purchase a stereo camera module, like the IMX219-83 Stereo Camera?
Thanks.
It depends on what you’re trying to achieve. With a single camera, you can use a depth estimation model to infer depth, but the accuracy often depends heavily on the model and training data. A stereo camera, on the other hand, can directly compute depth from geometry, usually giving more precise and reliable results. If you need rough depth information and want to keep hardware simple, a single camera may be fine. But if your application requires accuracy for instance for a robotic application, a stereo camera is often the better choice.
Thanks for the response and suggestion! Since I’m planning to use this setup for a robotic application, it seems like a stereo camera would be the better choice for more precise depth estimation. I wanted to ask if there are any specific camera modules supported by the Hailo AI software? For instance, would the IMX219-83 Stereo Camera work with the Hailo-8L, or is there another camera that you would recommend?
That’s not really something the Hailo AI software stack is concerned with. It doesn’t directly support or restrict specific camera modules. What matters is whether your host platform and OS provide driver support for the sensor (e.g. via MIPI-CSI or USB). Once the camera is exposed to the system e.g. via V4L2 (Video for Linux version 2), you can grab frames with the framework of your choice e.g. GStreamer and feed them into the model running on the Hailo device.
Thanks for the detailed explanation! I really appreciate it.