Posts made by andyyuen

@ulver Thanks for your kind comment. Yes, I hope my work will generate more interest on the LLM Module Kit. Looking forward to seeing announcements of advanced projects on the LLM Module Kit here soon.

In my previous project (A Web UI for the LLM Module Kit Voice Assistant), I showed you how to add a web user interface to interact with a LLM model using Arduino on a M5Stack core controller stacked on top of the LLM Module Kit.

This time I am showing you how to build a web UI running natively on the LLM Module Kit without a M5Stack controller. In other words, you can do it just using the LLM Module Kit alone. And it will have the same web UI which you saw in my previous project.

The programming environment I use this time is not Arduino. I am using Python3 and its flask and flask-sock modules. The web UI looks exactly the same as before with minor enhancements: the 'clear' and 'submit' buttons are now disabled when you click on 'submit' and re-eanabled when the LLM completes its response.

To access the webui, just point your browser to: http://llmModuleKitIPAddress:8080
Of course you have to run the webui.py first.

Source code and instructions can be found in my Github repository.

Enjoy!

The Voice Assistant Arduino example allows you to interact with the LLM Module Kit using only speech:

• It uses KeyWord Spotting (kws) to detect a spoken Wake word
• Once detected, it triggers the Automatic Speech Recognition (asr) module to convert your question for the LLM from speech to text
• It then submits the converted text as input for LLM inference
• And outputs the inference results as speech using the Text To Speech (tts) module.

A web UI complements the the Voice Assistant in the following use cases:

• for a long answer to a question, it takes the Voice Assistant a long time to go through it using speech. It will be much quicker if the answer is displayed as text but not on the small screen of the M5Stack Core processor
• I want to copy all or part of the answer and use it somewhere else
• the web UI and voice assistant can be used concurrently

My Youtube video shows the web and voice UI working independently and concurrently.

Web UI by itseld

Web and Voice UI Concurrently

The source code can be found at Github.

Information on the M5Stack LLM Module Kit can be found here.

And info on the M5Stack Gray here .

Enjoy!

Thermal cameras serve many purposes eg, one can use a thermal camera to monitor equipment in manufacturing and other industries. The thermal camera generates an event when the equipment gets too hot. Staff receiving the notification can use an app running in a container on the edge to receive the real-time thermal image stream and temperature of the equipment eg, to identify which part of the equipment is overheating.

The problem with thermal cameras is that they are expensive costing thousands of dollars. The less expensive ones usually have low resolution. We are talking about 8 X 8 pixels thermal images a few years back and more recently 32 X 24 pixels eg, the M5stack T-lite thermal camera. You may not be able to tell what you are looking at from the low resolution thermal images as illustrated in my YouTube video.

This project solves the problem at a low cost by augmenting the low resolution thermal images with more details provided by a second visible light camera. Imaging that in an Edge environment, you can now watch, in real-time, the augmented thermal image stream using a Python web application running in a container on Podman, MicroShift or OpenShift. Or you can just run the Python application on your notebook without using a container.

The instructions on how to build such an augmented thermal camera, the application architecture, container image creation and source code can be found at https://github.com/AndyYuen/augmented-thermal-camera.

Here is an animated preview of what to expect from an augmented thermal image compared to a raw thermal image: