Here is a smart sensor that brings LLM based vision to your projects with only wires. One of the challenges of LLMs is the amount of time it takes to generate a response. Unisense combines a motion sensor with a camera to only send an image to the model when it needs to see what is happening. It can recognize any object that can be described with words.
Must see 🚁 racing & photography drone courses
The camera module can rotate 120 degrees up or down. Once the motion sensor is triggered, the image is sent to the model, which sends a custom token to the microcontroller. Unisense supports most major inference platforms right out of the box, including OpenAI. The makers recommend MiniCPM Llama3 v2.5 for local inference.
[HT]
*Our articles may contain aff links. As an Amazon Associate we earn from qualifying purchases. Please read our disclaimer on how we fund this site.
