Hand labelling data for image classification mac os

Hand labelling images for my grassland ai

[@bluegtturbo]

Wow this is such a coincidence! I'm planning to develop a sprayer robot mainly for spraying hedges in the field, but have also considered the use of AI to develop an autonomous sprayer to detect and spray docks. I was thinking of using a rover instead of a drone, but I'm sure many of the same challenges will apply)

[@TheRealMrLaserCutter]

bluegtturbo rover is my plan too, wheelchair motors and Motorcross wheels. :)

[@bluegtturbo]

Exactly my thinking too. I'd use a 32A sabertooth R/C motor controller + Raspberry PI to drive the wheelchair motors. Essentially I'm on the lookout for an old electric wheelchair to use for a chassis, and have the option of both remote control and autonomous control using a webcam and a range of motion sensors and GPS linked to the raspberry pi. The remote control part seems far more straightforward. The AI system to activate the sprayer solenoid could use Google's Tensorflow, but I'd prefer a system that doesn't rely on an internet connection. My approach will be to start off working on individual elements of the overall architecture. For example, first develop an image recognition system to activate the sprayer solenoid, maybe just use that on the back of a tractor initially to test and fine tune it. If and when it works satisfactorily , acquire an electric wheelchair and attach it to that, and develop some level of autonomy.

[@TheRealMrLaserCutter]

Tensorflow can be run locally on a laptop or even raspberry pi, future pi's will be better at this though. You can see my project on the fast.ai forum under the deep learning section. It's called grassland spryaer. I was thinking about setting up a group to work on this project.

[@TheRealMrLaserCutter]

bluegtturbo I think we all need to try things and see what works, then take the best of everyones experiments and move forward.

[@bluegtturbo]

I'm with you on that CNC - a collaborative approach will be fun and hopefully more productive. Thank you for clarifying regarding Tensorflow - that is good news!