I tried using 'Doshisha Miscon AI' which judges the probability of becoming Miss Doshisha from the face photograph

Miss Doshisha (Miss Campus Doshisha) is selected from six candidates of female students enrolled in Doshisha University, and it can be said that it is a very narrow gate considering that the number of female students is stronger I will. Mr. Tango Kaya said that he developed " Doshisha MISCON AI " to calculate the probability of becoming Miss Doshisha from the face photograph under the idea of ​​"It would be fun if anyone can easily participate in Miss Doshisha? about. I actually tried using it.

Doshisha Miscon AI
https://miss-ai-doshisha.herokuapp.com/

Doshisha MISCON AI is AI using convolution neural network (CNN). This AI uses 2000 images of participants nominated by Miss Doshisha in the past 3 years, and if you load a face picture into Doshisha Miscon AI, "Probability of being chosen as a finalist 6" and "Grand Prix The probability of becoming "will be displayed.

◆ Actually trying to use - Access to Doshisha Miscon AI quickly.

Scroll down and click "Select file" button.

Select face photo of editorial staff (man) ......

Click "Judge" button.

Then, it cuts out the face part automatically from the uploaded image file, and the person who is similar in past mistake Doshisha final list, and the probability that it becomes "finalist by applying to Miss Doshisha", "Miss Doshisha Grand Prix Probability to shine "is displayed. In the case of facial photographs of editorial staff, AI is evaluated similar to Mr. Yukiyama Yokoyama who was selected as the finalist of 2015, the probability of becoming Miss Doshisha finalist is 18.83%, the probability of Miss Doshisha Grand Prix shining is 0.01% It was evaluated. I feel kindness in the part that is evaluated as 0.01% instead of 0%.

Next, I will use a female face photo instead of a male. In this example, let's load the following faces generated by "idle generation AI" which automatically generates fictitious idols .

By Data Corporation

Then it is judged that it is similar to Miki Sugiura of Grand Prix of 2017, Minami Aoki of Grand Prix of 2016 and Maki Yonekura of finalists of 2015, and the probability of becoming a finalist becomes 100%, becoming a Grand Prix The probability was determined to be 72.4%. When it comes to the face of the idle level, the evaluation seems to be quite high.

Doshisha MISCON AI is a "joke application" that AI judges based only on whether it is similar to the past Grand Prix winner or the face of the finalist. People who actually try to challenge MISCON should never take this result.