Method for editing image based on artificial intelligent and artificial device
A method for editing an image based on artificial intelligence and an artificial intelligence device are disclosed. In the method for editing an image based on artificial intelligence, a recommended photographic composition is created by applying an image acquired by a camera to a pre-trained composition recommendation model, and the image is corrected based on the created recommended photographic composition, whereby it is possible to capture and edit an image at the same time without the trouble of correcting the image after capturing it. One or more of an artificial intelligence device of the present disclosure can be associated with artificial intelligence modules, drones (unmanned aerial vehicles (UAVs)), robots, augmented reality (AR) devices, virtual reality (VR) devices, devices related to 5G service, etc.
1. A method for editing an image based on artificial intelligence, the method comprising:
acquiring a preliminary image by a camera;
applying the preliminary image to a pre-trained composition recommendation model;
creating a recommended photographic composition based on an output from the composition recommendation model;
correcting the preliminary image based on the recommended photographic composition,
wherein the composition recommendation model is an artificial neural network model that is trained by setting up first photographic composition data and second photographic composition data as a training data set,
the first photographic composition data being extracted from the preliminary image, and the second photographic composition data being extracted from the from a recommended image;
obtaining a user's history of selecting recommended images and the user's image editing history;
updating the training data set based on the user's history of selecting recommended images and the user's image editing history; and
training the composition recommendation model based on the updated training data set.
2. The method of claim 1 , wherein the recommended image is a professional photographer's photographic image.
3. The method of claim 1 , further comprising:
identifying objects included in the preliminary image; and
obtaining photographic composition data based on positions of the identified objects.
4. The method of claim 3 , wherein the photographic composition data comprises at least one among types of objects, a number of objects, and a main object.
5. The method of claim 4 , further comprising determining the main object.
6. The method of claim 5 , wherein the determining of the main object comprises identifying a person pre-registered on an artificial intelligence device,
wherein, if the pre-registered person exists, the pre-registered person is set as the main object.
7. The method of claim 5 , wherein the determining of the main object comprises determining the main object based on a proportion of each of a plurality of objects included in the preliminary image.
8. The method of claim 5 , wherein the determining of the main object comprises determining the main object based on the camera's focusing information.
9. The method of claim 1 , further comprising, if there is at least one recommended image, displaying the recommended image on a display.
10. The method of claim 1 , wherein the applying of the preliminary image to a pre-trained composition recommendation model further comprises:
transmitting the preliminary image to an external server connected to the artificial intelligence device to apply the same to a composition recommendation model stored in the external server, and
receiving a recommended photographic composition created based on an output from the composition recommendation model stored in the external server.
11. An artificial intelligence device comprising:
a memory that stores a trained composition recommendation model and a training data set for the composition recommendation model;
a transceiver that communicates with an external server;
a camera that captures a preliminary image; and
a processor that applies the preliminary image to the composition recommendation model, creates a recommended photographic composition based on an output from the composition recommendation model, and corrects the preliminary image based on the recommended photographic composition,
wherein the composition recommendation model is an artificial neural network model that is trained by setting up first photographic composition data and second photographic composition data as a training data set,
wherein the first photographic composition data is extracted from the preliminary image, and the second photographic composition data is extracted from a recommended image, and
wherein the processor obtains a user's history of selecting recommended images and the user's image editing history, updates the training data set based on the user's history of selecting recommended images and the user's image editing history, and trains the composition recommendation model based on the updated training data set.
12. The artificial intelligence device of claim 11 , wherein the processor identifies objects included in the preliminary image, and obtains photographic composition data based on positions of the identified objects.
13. The artificial intelligence device of claim 12 , wherein the photographic composition data comprises at least one among types of objects, a number of objects, and a main object.
14. The artificial intelligence device of claim 13 , wherein the processor determines the main object.
15. The artificial intelligence device of claim 14 , wherein a person pre-registered on the artificial intelligence device is set as the main object.
16. The artificial intelligence device of claim 14 , wherein the main object is determined based on a proportion of each of a plurality of objects included in the preliminary image.