Systems, methods, and media for colorizing grayscale images

One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

• capture a first grayscale image by a first camera of a device and a second grayscale image by a second camera of the device, wherein the first and second grayscale images are stereo images that are captured at a same time;
• access the first grayscale image and the second grayscale image that are captured by the first camera and second camera, respectively;
• generate, using a machine-learning model, a first color image corresponding to the first grayscale image and a second color image corresponding to the second grayscale image, wherein the first and second color images include color inconsistencies;
• generate affinity information, for resolving the color inconsistencies in the first and second color images, based on comparison of pixels between the first grayscale image and the second grayscale image, the affinity information identifying relationships between pixels of the first grayscale image and pixels of the second grayscale image;
• modify the first color image and the second color image, which are generated using the machine-learning model, based on the relationships between the pixels of the first grayscale image and the pixels of the second grayscale image, wherein the modification resolves the color inconsistencies in the first color image and the second color image;
• and
• generate a first visual output based on the modified first color image and a second visual output based on the modified second color image.

A system comprising:

• one or more processors;
• and
• one or more computer-readable non-transitory storage media coupled to one or more of the processors and comprising instructions operable when executed by one or more of the processors to cause the system to:
• capture a first grayscale image by a first camera of a device and a second grayscale image by a second camera of the device, wherein the first and second grayscale images are stereo images that are captured at a same time;
• access the first grayscale image and the second grayscale image that are captured by the first camera and second camera, respectively;
• generate, using a machine-learning model, a first color image corresponding to the first grayscale image and a second color image corresponding to the second grayscale image, wherein the first and second color images include color inconsistencies;
• generate affinity information, for resolving the color inconsistencies in the first and second color images, based on comparison of pixels between the first grayscale image and the second grayscale image, the affinity information identifying relationships between pixels of the first grayscale image and pixels of the second grayscale image;
• modify the first color image and the second color image, which are generated using the machine-learning model, based on the relationships between the pixels of the first grayscale image and the pixels of the second grayscale image, wherein the modification resolves the color inconsistencies in the first color image and the second color image;
• and
• generate a first visual output based on the modified first color image and a second visual output based on the modified second color image.