问题
Like, is this possible?
Basic idea is to make a robot that can see the yard, recognize a dandelion flower, and then go murder the stinking monster where it sits.
The robot is one thing. Telling it whut to do is another. Figuring out where you want to tell the robot to go is the issue. Like, find a yellow thing in the yard, check if it's maybe one of the devil spawn, get closer, then make sure it IS the evil target so we don't kill moms purdy flowers or the neighbors dog.
Which MS CS tool would you use and how might it be implemented?
回答1:
You can use Computer Vision (https://azure.microsoft.com/en-us/services/cognitive-services/computer-vision/) to start with, but you may find that it detects "flower" but not necessarily "dandelion". If the recognition is not as accurate as you need it to be then you can use Custom Vision (https://azure.microsoft.com/en-us/services/cognitive-services/custom-vision-service/) where you can train the API to recognize the specific objects that you are interested in.
来源：https://stackoverflow.com/questions/44530182/discussion-how-to-use-visual-recognition-to-locate-a-dandelion-weed-or-flower