In a previous post Raspberry Pi AI Camera post I highlighted its ability to run on all Raspberry Pi boards due to its built-in processing chip and memory. All you need to do is simply connect the AI Camera via the CSI interface.
Now let's take a look at the kind of applications and projects this camera can do.
Models
Firstly, there are two built-in AI models, namely MobileNet SSD and PoseNet.
MobileNet SSD is applicable for real-time object detection applications, while PoseNet is trained for pose identification.
To start just run the rpicam-hello with your choice of model. The model will be transferred to the camera module over the CSI connection (a process handled by the on-board RP2040) and loaded into 8MB of on-device RAM. The model stays loaded and constantly proceses images in real-time.
However, there is more. For example Raspberrry Pi have their Github libcamera repository, with three Python code examples for image classification, object detection and segmentation.
These code examples work with a range of models including EfficientNET v2, MobileViT XS, ResNet18, SqueezeNet and YOLOv8n. YOLOv8 can run at the maximum 640×640 resolution with others are limited to lower resolutions.
Take a look at the full libcamera documentation to discover how to get started.
Projects
As for project ideas they are almost limitless.Here are a few examples:
• Smart Doorbells
• Smart Room/Baby Monitors
• Gesture Controled Devices
• Smarter Robots
• Pet or Wildlife Observation
But, of course, the fun is in coming up with your own ideas - either by taking a different approach to an existing idea, or coming up with a brand new one.
No comments:
Post a Comment