INCREMENTAL MAP BUILDING USING LEARNABLE FEATURES AND DESCRIPTORS
A method for labeling keypoints includes labeling, via a keypoint model, a first set of keypoints in a first image associated with a three-dimensional (3D) map of an environment, the first image having been captured during a first time period. The method also includes labeling, via the keypoint model, a second set of keypoints in a second image associated with the 3D map, the second image having been captured during a second time period. The method further includes updating, via the keypoint model, one of more of the first set of keypoints based on labeling the second set of keypoints.
1 . A method for labeling keypoints, comprising:
labeling, via a keypoint model, a first set of keypoints in a first image associated with a three-dimensional (3D) map of an environment, the first image having been captured during a first time period;
labeling, via the keypoint model, a second set of keypoints in a second image associated with the 3D map, the second image having been captured during a second time period; and
updating, via the keypoint model, one of more of the first set of keypoints based on labeling the second set of keypoints.
2 . The method of claim 1 , wherein the keypoint model is trained to label keypoints for two-dimensional (2D)-to-3D keypoint matching.
3 . The method of claim 1 , wherein:
the first image is one of a first group of images associated with the 3D map;
the second image is one of the second group of images associated with the 3D map;
one or more images of the second group of images are the same as one or more images of the first group of images; and
the first group of images and the second group of images are two-dimensional (2D) images.
4 . The method of claim 3 , wherein the first group of images and the second group of images are captured via a sensor associated with a vehicle.
5 . The method of claim 4 , wherein the keypoint model is remotely located from the vehicle.
6 . The method of claim 4 , further comprising:
capturing, via the sensor, a 2D query image of the environment;
identifying a target image, from the first group of images and the second group of images, comprising a set of target keypoints that match a set of query keypoints of the 2D query image;
determining a current location within the 3D map based on identifying the target image; and
autonomously or semi-autonomously navigating through the environment based on determining the current location, wherein the keypoint model and the 3D map are stored at the vehicle.
7 . The method of claim 1 , wherein each of the first set of keypoints and the second set of keypoints are associated with features of the environment.
8 . An apparatus for labeling keypoints, comprising:
one or more processors; and
one or more memories coupled with the one or more processors and storing processor-executable code that, when executed by the one or more processors, is configured to cause the apparatus to:
label, via a keypoint model, a first set of keypoints in a first image associated with a three-dimensional (3D) map of an environment, the first image having been captured during a first time period;
label, via the keypoint model, a second set of keypoints in a second image associated with the 3D map, the second image having been captured during a second time period; and
update, via the keypoint model, one of more of the first set of keypoints based on labeling the second set of keypoints.
9 . The apparatus of claim 8 , wherein the keypoint model is trained to label keypoints for two-dimensional (2D)-to-3D keypoint matching.
10 . The apparatus of claim 8 , wherein:
the first image is one of a first group of images associated with the 3D map;
the second image is one of the second group of images associated with the 3D map;
one or more images of the second group of images are the same as one or more images of the first group of images; and
the first group of images and the second group of images are two-dimensional (2D) images.
11 . The apparatus of claim 10 , wherein the first group of images and the second group of images are captured via a sensor associated with a vehicle.
12 . The apparatus of claim 11 , wherein the keypoint model is remotely located from the vehicle.
13 . The apparatus of claim 11 , wherein execution of processor-executable code further causes the apparatus to:
capture, via the sensor, a 2D query image of the environment;
identify a target image, from the first group of images and the second group of images, comprising a set of target keypoints that match a set of query keypoints of the 2D query image;
determine a current location within the 3D map based on identifying the target image; and
autonomously or semi-autonomously navigate through the environment based on determining the current location, wherein the keypoint model and the 3D map are stored at the vehicle.
14 . The apparatus of claim 8 , wherein each of the first set of keypoints and the second set of keypoints are associated with features of the environment.
15 . A non-transitory computer-readable medium having program code recorded thereon for labeling keypoints, the program code executed by one or more processors and comprising:
program code to label, via a keypoint model, a first set of keypoints in a first image associated with a three-dimensional (3D) map of an environment, the first image having been captured during a first time period;
program code to label, via the keypoint model, a second set of keypoints in a second image associated with the 3D map, the second image having been captured during a second time period; and
program code to update, via the keypoint model, one of more of the first set of keypoints based on labeling the second set of keypoints.
16 . The non-transitory computer-readable medium of claim 15 , wherein the keypoint model is trained to label keypoints for two-dimensional (2D)-to-3D keypoint matching.
17 . The non-transitory computer-readable medium of claim 15 , wherein:
the first image is one of a first group of images associated with the 3D map;
the second image is one of the second group of images associated with the 3D map;
one or more images of the second group of images are the same as one or more images of the first group of images; and
the first group of images and the second group of images are two-dimensional (2D) images.
18 . The non-transitory computer-readable medium of claim 17 , wherein the first group of images and the second group of images are captured via a sensor associated with a vehicle.
19 . The non-transitory computer-readable medium of claim 18 , wherein the keypoint model is remotely located from the vehicle.
20 . The non-transitory computer-readable medium of claim 18 , wherein the program code further comprises:
program code to capture, via the sensor, a 2D query image of the environment;
program code to identify a target image, from the first group of images and the second group of images, comprising a set of target keypoints that match a set of query keypoints of the 2D query image;
program code to determine a current location within the 3D map based on identifying the target image; and
program code to autonomously or semi-autonomously navigate through the environment based on determining the current location, wherein the keypoint model and the 3D map are stored at the vehicle.