Brought to you by: Finn
Badly Drawn Bunnies
A new computer program can recognize semantic sketches — rough,
cartoony depictions that actually look nothing like the animal they are
supposed to represent.
Mathias Eitz, James Hays and Marc Alexa
12 comments:
Wow. To think that a computer can recognize out simple sketches is ... amazing. Sure, it misses sometimes, but don't we all? Have you seen the time when a Jeopardy! winner played the game against a robot? The robot won. I think that we should do that. The scientists should play a game of Pictionary. There would be a human and a phone with this app on it. The scientist would draw something on the app. Then, they would see who recognized it correctly, the human or the computer. I think that the human would win.
Out of 10 times how many times does it fail? The article says that it can recognize badly drawn things. What about good drawings? Can it recognize those? What would it recognize a abstract drawing as?
Overall,I believe that this was a very interesting article. Thanks, Finn. There was another article like this, about robots learning to identify things. If this continues, robots may become just like humans, able to think and feel emotion. Then they will almost be humans. For that is what makes us us. Our ability to think and feel makes us human. If robots are able to do this, then they are human.
Amazing! I definitely agree with Emilynne. It was very interesting to compare a human's accuracy with badly drawn cartoons to a robot's accuracy. The robot failed to reconize the drawings, but people can not reconize it either.
So how can the robot be changed so that it can correctly guess the cartoon? If the robot went to an art museum, would it reconize the more abstract and vague paintings? Is this the start to a new age in the world where robots and humans can work together and do the same things?
This was a great article! It made me think about it and want to know more so that I could find out what this could be in twenty years or so.
I think that this program is amazing, agreeing with Alyssa and Emilynne. The idea that our horrible stick figures and dog and cats can be identified by a computer is really cool. But, before the programmers release this as an app or program on iPads and phones, they should make sure that the program identifies objects correctly 70% of the time or more. 55% is only slightly more than half, and would be easy to make malfunction if you are bad at drawing. Even though this program is, apparently, made if you aren't good at it, it seems like that isn't what would really happen. This program would have to be put into serious thought before releasing.
I also think it is amazing that a computer can recognize bad drawings. I definitely agree with Emilynne about the Pictionary game with the scientists and the phone.
The one thing I just don't understand is why the computer thought the dogs were a mosquito and a mermaid.
This article proved that computers are not that smart. None of the dog drawings were recognized. The one predicted "panda" was actually really obviously a dog. The particular computer is not very smart.
I actually think that a game of pictionary might not work. We would have to have a person drawing something with the computer guessing and another person drawing with another person guessing. The computer might have a faster way to recognize these things, but a human brain is probably more accurate. I agree with Alyssa. If the robot went to a museum and saw a Pablo Picasso or a splatter piece, what would it recognize it as? I think it would count it as bad, even though it is perfectly good art.
In all, I think that robots and humans could work together to make a smarter world, but maybe just not for this purpose.
I think it was really cool how they created the program to detect really bad drawings. I think the article should have talked about how they made the program.
I think that it is cool that they have created a program that detects bad drawings, but they need to improve the program drastically. To the computer, a dog looks like everything. Penguin, mosquito, and mermaid were the funniest guesses in my opinion. The article said the computer recognized 56% of drawings, compared to a human's 73%. That is saying that they have created a program that cannot do what a human can do. The point of computers is to do thing's that human can't. And, these kinds of things have already been tried. Like the article said, IBM made Watson to be on Jeopardy, and it was much better than what this computer was guessing. Also, the Watson one was much more useful. What's the point of having a game of pictionary with the computer? I would have to say that this is a program that just does not need to be made, even though it is pretty cool.
I agree with Emilynne and Alyssa. It is very interesting that a computer can try to detect what you are drawing. Eve though the robot failed to pass the challenge humans may not be able to recognize these drawings either.
I think this article proves that robots are very smart. I am excited to see how smart robots are in the future. I am also confused on why the computer thought dogs were a mosquito, or a mermaid.
I think that this article does not exactly prove the intelligence of robots. I mean, to agree with Nick, the robot only recognize 56% of the drawings! The whole point of the robot is to do some pretty incredible things that humans can't do. This article proved that this robot was not the "smartest" out of robots.
I think that this article could have been more elaborate on HOW they made this program. But I thought it was cool that the people made this program to detect bad drawings. I didn't really enjoy this article very much.
I think it's cool that a computer can recognize sketches even if they are drawn badly. But it only senses images 56% of the time.
Before we start really getting into this we need to find something as, if not more precise than the human eye. 56% of the time is not nearly accurate enough. Even though this is a breakthrough, it cannot be thought of as accurate until we really have accurate technology that can sense greater than or equal to 73%.
This article proves that we are improving technology, however this cannot be thought of as a huge success until we measure this, as Emilynne suggested against a human eye. We have to make computers able to really identify them before we can move on.
Overall, this article was very interesting. I think that we should keep doing research to improve the field of technology. If we keep working, I bet we could have an accurate device in the next twenty years.
Post a Comment