Create AI using tinyML seeed Wio Terminal D51R

During Bali Fab Fest 17, I am join to Workshop "Mini-Hackathon on Using AI Vision and IoT for Your Own Sustainability! Projects". In this workshop, we learn about how to develop Vision AI using tiny microcontroller using Seeed studio product sensecap k1100, this package has modular sensor such as soil sensor, temperature and humidity, eC02 Gas sensor and also Lora module to connecting into lorawan gateway so we can monitoring by real-time sensor data that attached.

in first day we try to compile every sensor on it and this kit are make me shock, we didn't do coding anything to do that

so LoRa it's a one of protocol communication for IoT devices, it has a different spectrum frequency radio every country, and for frequency in Indonesia is 923. Benefits of using LoRa is range of the connection distance is so wide, commonly for using standard antenna it can coverage for more than 10 km, and also it took low-power consumption to device, so it can possible by using the battery and it can be turn it on until 2 years or more.

for day 2, we learn about how to create AI vision using little seeed wio terminal, we all know to create AI training it need a high specification devices for create library, but by using roboflow we don't need to get each image itself to be trained data

to be make seeed camera will understand so we need to download that YOLOv5 data from robloflow and trained the data using colab.research.google.com

and after data trained, next the data converted into google tensorflow lite by using python language, and for final, convert into uf2 so the camera can detect the program using the new algorithm.

as you can see, at the top of the image, if we make direct to apple image, now it can detect apple by the green rectangular and show number of percentage value of apple match image.