Depth estimation method and apparatus, method and apparatus for training depth estimation model, and system

1 . A depth estimation method performed by a depth estimation device, comprising: obtaining a first color image; inputting the first color image into a first depth estimation model, to obtain a first intermediate depth image, wherein the first depth estimation model is obtained through training based on a training color image and an intermediate depth image corresponding to the training color image; and inputting the first color image and the first intermediate depth image into a second depth estimation model, to obtain a first target depth image, wherein the second depth estimation model is obtained through training based on the training color image and a target depth image corresponding to the training color image, wherein the first depth estimation model and the second depth estimation model are neural network models, wherein the first target depth image is a refined image obtained based on the first intermediate depth image, wherein inputting the first color image and the first intermediate depth image into the second depth estimation model to obtain the first target depth image comprises: obtaining a common edge image based on the first color image and the first intermediate depth image, wherein the common edge image represents a common edge part of the first color image and the first intermediate depth image; and obtaining the first target depth image based on the common edge image and the first intermediate depth image.

2 . The method according to claim 1 , wherein the step of obtaining the common edge image comprises: splicing the first color image and the first intermediate depth image, to obtain an input matrix; performing dimension reduction on the input matrix, by separately inputting the input matrix into a channel attention module to obtain a first matrix and into a spatial attention module to obtain a second matrix; and obtaining the common edge image based on the first matrix and the second matrix.

3 . The method according to claim 1 , wherein the step of inputting the first color image into the first depth estimation model to obtain the first intermediate depth image comprises: obtaining a depth feature vector of the first color image based on the first color image; performing a split operation on the depth feature vector to obtain a first part of the depth feature vector and a second part of the depth feature vector; inputting the first part and the second part into respective multi-scale feature paths to obtain a first depth feature vector and a second depth feature vector; obtaining a first target feature vector based on the first depth feature vector and a first weight value, and obtaining a second target feature vector based on the second depth feature vector and a second weight value; and obtaining the first intermediate depth image based on the first target feature vector and the second target feature vector by splicing the first target feature vector and the second target feature vector.

4 . The method according to claim 1 , wherein the training color image comprises a left view or a right view of a vehicle.

5 . The method according to claim 4 , wherein the training color image is shot by a camera disposed on the vehicle.

6 . The method according to claim 5 , wherein the camera is a binocular camera.

7 . The method according to claim 1 , further comprising:

calculating gradient information of the first depth estimation model and the second depth estimation model based on the first color image and the first target depth image; and

sending the gradient information to a first device, wherein the gradient information is used by the first device to obtain an updated first depth estimation model and an updated second depth estimation model through training.

8 . The method according to claim 7 , further comprising:

receiving the updated first depth estimation model and the updated second depth estimation model from the first device.

9 . The method according to claim 7 , further comprising:

obtaining the updated first depth estimation model and the updated second depth estimation model through training based on the gradient information.

10 . A method for training a depth estimation model, comprising: processing a color image by using a first depth estimation model, to obtain an intermediate depth image of the color image; processing the color image and the intermediate depth image of the color image by using a second depth estimation model, to obtain a target depth image of the color image; training the first depth estimation model based on the color image and the intermediate depth image of the color image, to obtain an updated first depth estimation model; and training the second depth estimation model based on the color image and the target depth image of the color image, to obtain an updated second depth estimation model, wherein the first depth estimation model and the second depth estimation model are neural networks, wherein the target depth image is a refined image obtained based on the intermediate depth image, wherein the updated second depth estimation model is configured to: obtain a common edge image based on the first color image and the first intermediate depth image, wherein the common edge image represents a common edge part of the first color image and the first intermediate depth image; and obtain the first target depth image based on the common edge image and the first intermediate depth image.

11 . The training method according to claim 10 , wherein the step of obtaining the common edge image comprises: splicing the first color image and the first intermediate depth image, to obtain an input matrix; performing dimension reduction on the input matrix, by separately inputting the input matrix into a channel attention module to obtain a first matrix and into a spatial attention module to obtain a second matrix; and obtaining the common edge image based on the first matrix and the second matrix.

12 . The training method according claim 10 , wherein the updated first depth estimation model is configured to: obtain a depth feature vector of the first color image based on the first color image; perform a split operation on the depth feature vector to obtain a first part of the depth feature vector and a second part of the depth feature vector; input the first part and the second part into respective multi-scale feature paths to obtain a first depth feature vector and a second depth feature vector; obtain a first target feature vector based on the first depth feature vector and a first weight value, and obtain a second target feature vector based on the second depth feature vector and a second weight value; and obtain the first intermediate depth image based on the first target feature vector and the second target feature vector by splicing the first target feature vector and the second target feature vector.

13 . The training method according to claim 10 , wherein the color image comprises a left view or a right view of a vehicle.

14 . The training method according to claim 10 , further comprising: receiving gradient information of the first depth estimation model and the second depth estimation model, wherein the gradient information is obtained by a device through calculation based on the first color image and the first target depth image; and obtaining the updated first depth estimation model and the updated second depth estimation model through training based on the gradient information.

15 . The training method according to claim 14 , further comprising:

sending the updated first depth estimation model and the updated second depth estimation model.

16 . A depth estimation apparatus comprising: a data interface; a memory storing executable instructions; and a processor configured to execute the executable instructions to perform operations of: obtaining, through the data interface, a first color image; inputting the first color image into a first depth estimation model to obtain a first intermediate depth image, wherein the first depth estimation model is obtained through training based on a training color image and an intermediate depth image corresponding to the training color image; and inputting the first color image and the first intermediate depth image into a second depth estimation model to obtain a first target depth image, wherein the second depth estimation model is obtained through training based on the training color image and a target depth image corresponding to the color image, wherein the first depth estimation model and the second depth estimation model are neural network models, wherein the first target depth image is a refined image obtained based on the first intermediate depth image, wherein inputting the first color image and the first intermediate depth image into the second depth estimation model to obtain the first target depth image comprises: obtaining a common edge image based on the first color image and the first intermediate depth image, wherein the common edge image represents a common edge part of the first color image and the first intermediate depth image; and obtaining the first target depth image based on the common edge image and the first intermediate depth image.

17 . The depth estimation apparatus according to claim 16 , wherein the operation of obtaining the common edge image comprises: splicing the first color image and the first intermediate depth image, to obtain an input matrix; performing dimension reduction on the input matrix, by separately inputting the input matrix into a channel attention module to obtain a first matrix and into a spatial attention module to obtain a second matrix; and obtaining the common edge image based on the first matrix and the second matrix.