"CaptionBot" which can experience Microsoft's image recognition function automatically analyzing images and explaining by words

The service that the computer analyzes the uploaded image and the image of the specified URL and judges what is shown and explains it in natural language "CaptionBot"Is provided. I tried the ability of a site that can experience a part of the image recognition function said to be approaching human judgment ability.

CaptionBot - For pictures worth the thousand words
https://www.captionbot.ai/


CaptionBot is a service that you can experience image recognition by accessing from the browser and uploading an image or specifying the URL of the image.


In this way, after specifying an image in Explorer, click "Open" to upload ... ...


The image was analyzed in a few seconds, and the result was displayed on the image. It seems that this image was a bit difficult to judge with this image, "I do not have confidence, but I think that it is a hand holding a cup", it is quite a bad answer. In the part of "How did I do?" Written in the lower left of the screen (the result was how?), It is now possible to evaluate the result by the number of stars.


It is an image of a fighter plane, "I think that it is a fighter jet flying in blue sky," this is a big right answer.


When it is an image of a cat spotted in a box, "I think that it is a cat that tied a necktie", the only part of the cat is correct.


Chilled ChamponThe image of "salads were cooked with one dish and water bottle", not far from being hit. I think it is difficult for human beings, but there are things that noodles are not visible, so it seems impossible to reach the keyword "Champion" as expected.


An explanation of the meaning that "Champagne raised chopsticks with chopsticks," which means "I cooked pasta and vegetables dishes up". Where you are concerned whether the keyword "chopstick (chopstick)" appeared when you used a slightly more pictures.


SpaceX's "Falcon 9" rocket successfully landed on an offshore drone shipThen, "An ocean-sailing ship" answer.


Pictures of drone and iPhone and iPad placed on top of the concrete block seemed to be a little difficult, "I do not have confidence, but I think that it is a bird sitting on a box." Although it is out of the way, I think the thought process that came to judge the drones as "birds" is very anxious.


In addition, CaptionBot'sComment pageAccording to this serviceMicrosoft Cognitive ServicesIt is said that it is provided using the recognition function of Microsoft Cognitive Service. It is designed to guide answers using Computer Vision API and Emotion API, Bing Image API, and Natural Language which generates natural language.


Also, CaptionBot is a brotherly position service such as "Age guessing function" How Old Do I Look?

Microsoft official machine learning site "How Old Do I Look?" Which guesses age from images - GIGAZINE