How to recognize the color and shape of the case in the image of a watch?

Question

Images

There is no hope for an answer, but let's try
. Tell me the idea of ​​\u200b\u200bimplementing such a task.

There are 40 thousand images of wristwatches. Several characteristics need to be determined from the image:
1) The shape of the case (Oval, barrel, circle, square, rectangle)
3) The color of the dial
4) The color of the strap


How can modern technologies solve such problems?
Neural networks or Yandex.Toloka?
Or are there easier options?

Answer 1

See how you can do it, make a simple site, the user logs in, he is given images and he fills in the parameters (color, shape). Next, we go to a site like wmmail, set a task there with an amount of, for example, $ 1 per 1000 images, but without auto-payment. It is important that each verified image is tied to a specific user, so that if someone cheats, you can do a reset, and to check, for example, view 10-20 random images and parameters filled in by the user, if you make a mistake somewhere, then don’t pay and reset everything filled in by the user. Well, krch schoolchildren will scatter this task for you in a day, they are ready to jump out of the window for $ 1.
As a result, you will complete the task and, based on this data, you will be able to train if such a task may arise in the future.

Answer 2

Neural networks. Easy to write in Python.