Usless notill
I looked online and the highest people get on seendling plant classification is 97.103% out of 12 different plant categories of 60000 images, which isn't very good.
I found another with plant disease detection based on photos of a leaf from a plant, and it one guy got 96.74% accuracy, which also isn't very good.
There is corn recognition where some people got within 0.00001 standard deviation on the softmax (basically 99.99999% accurate, but there are only ~2000 images so that is probably just overfitting and doesn't really count.
I tried to copy the first example on seedling classification and within like 4 epochs i got like 20%, but it runs so slow on my laptop that each epoch takes like a minute, and it would take 1000's to get up to the >95%.
if the precision application either by drones or whatever to the football field or the clovers on the field behind the track, it would need better than those for Mr. Perry or whoever to consider it practical or at least comparable to the accuracy of humans
both field sprayed like 2 times a year.
bruce_a
Comments