In the previous article [1], I mentioned the recognition of photos from live cameras that shoot mountains. And I have developed an app that can automatically receive it by e-mail only when a photo with a clear view of the top is obtained. In order to recognize photos, I created a model in which many photos of the mountain were classified into four categories and trained. There, I used MIT's Personal Image Classifier and Google's Teachable Machine2 and incorporated the trained model into the App Inventor app.
However, in general, mountains are often covered with clouds. For example, at Rishiri Mountain in northern Hokkaido, you can see the middle but rarely see the top. So, I checked the weather forecast, and on a sunny day, I ran this app continuously from the early morning and waited for a notification email. During that time, of course, I was doing another job because it is fully automatic notification!
Finally, good luck came. It happened after 12:00 that day. As shown in Figure 1, the mountain photo was judged to be “clear enough to see the summit” and I received an automatic email to inform it. I was very happy. Sure, I got a great photo that looks good up to the top of the mountain.
References
[1] Using MIT's Personal Image Classifier with App Inventor
http://sparse-dense.blogspot.com/2019/12/using-mits-personal-image-classifier.html
0 件のコメント:
コメントを投稿