Determining visual overlap of images by using box embeddings
An image matching system for determining visual overlaps between images by using box embeddings is described herein. The system receives two images depicting a 3D surface with different camera poses. The system inputs the images (or a crop of each image) into a machine learning model that outputs a box encoding for the first image and a box encoding for the second image. A box encoding includes parameters defining a box in an embedding space. Then the system determines an asymmetric overlap factor that measures asymmetric surface overlaps between the first image and the second image based on the box encodings. The asymmetric overlap factor includes an enclosure factor indicating how much surface from the first image is visible in the second image and a concentration factor indicating how much surface from the second image is visible in the first image.
1. A computer-implemented method comprising:
receiving a first image and a second image, the first and second images depicting a same 3D surface with different camera poses;
partitioning the first image into a first group of crops;
partitioning the second image into a second group of crops;
inputting each respective crop in the first group into a machine learning model, the machine learning model outputting a first box encoding of the respective crop in the first group;
inputting each respective crop in the second group into the machine learning model, the machine learning model outputting a second box encoding of the respective crop in the second group, each of the first and second box encodings including parameters defining a box in an embedding space; and
determining an asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings, the asymmetric overlap factor including an enclosure factor indicating how much surface from the first image is visible in the second image and a concentration factor indicating how much surface from the second image is visible in the first image.
2. The computer-implemented method of claim 1 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings comprises:
identifying intersections between the boxes for the crops in the second group and the boxes for the crops in the second group in the embedding space.
3. The computer-implemented method of claim 2 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining a total intersection by aggregating the identified intersections.
4. The computer-implemented method of claim 3 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining a total volume of the boxes for the crops in the second group based on the second box encodings.
5. The computer-implemented method of claim 4 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining the enclosure factor based on a ratio of the total intersection to the total volume of the boxes for the crops in the second group.
6. The computer-implemented method of claim 3 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining a total volume of the boxes for the crops in the first group based on the first box encodings.
7. The computer-implemented method of claim 6 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining the concentration factor based on a ratio of the total intersection to the total volume of the boxes for the crops in the first group.
8. A non-transitory computer-readable memory storing executable computer program instructions, the instructions executable to perform operations comprising:
receiving a first image and a second image, the first and second images depicting a same 3D surface with different camera poses;
partitioning the first image into a first group of crops;
partitioning the second image into a second group of crops;
inputting each respective crop in the first group into a machine learning model, the machine learning model outputting a first box encoding of the respective crop in the first group;
inputting each respective crop in the second group into the machine learning model, the machine learning model outputting a second box encoding of the respective crop in the second group, each of the first and second box encodings including parameters defining a box in an embedding space; and
determining an asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings, the asymmetric overlap factor including an enclosure factor indicating how much surface from the first image is visible in the second image and a concentration factor indicating how much surface from the second image is visible in the first image.
9. The non-transitory computer-readable memory of claim 8 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings comprises:
identifying intersections between the boxes for the crops in the second group and the boxes for the crops in the second group in the embedding space.
10. The non-transitory computer-readable memory of claim 9 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining a total intersection by aggregating the identified intersections.
11. The non-transitory computer-readable memory of claim 10 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining a total volume of the boxes for the crops in the second group based on the second box encodings.
12. The non-transitory computer-readable memory of claim 11 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining the enclosure factor based on a ratio of the total intersection to the total volume of the boxes for the crops in the second group.
13. The non-transitory computer-readable memory of claim 10 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining a total volume of the boxes for the crops in the first group based on the first box encodings.
14. The non-transitory computer-readable memory of claim 13 , wherein determining the asymmetric overlap factor measuring asymmetric surface overlaps between the first image and the second image based on the first and second box encodings further comprises:
determining the concentration factor based on a ratio of the total intersection to the total volume of the boxes for the crops in the first group.