i'm trying to realize a detector for diseased grape leaves, for this par of the project i'm just interested in detecting lets say, the percentage of diseased to healty leaves and/or place a flag where diseased leaves are detected.
Now, i'm pretty acquainted with ROS and i managed to run and tune some algorithms ( darknet / openai ) , but i never succesfully trained a network from scartch. So i'm a little baffled on how to proceed.
I also would like to minimized the labelling process, altough i'm not the one who will do it.
At first i tough of SegNets, i can take pictures of the vine and color healty / sick / not-leaf(?), but i'm afraid that would be as tedious as drawing rectangles?
But then, what about taking a lot of picutures of single leaves ( healty / non-healty ) would that be enough for a yolo-like detector? Or do that also need background for training?