Training a card type classifier with simulated card images
A computer model to identify a type of physical card is trained using simulated card images. The physical card may exist with various subtypes, some of which may not exist or be unavailable when the model is trained. To more robustly identify these subtypes, the training data set for the computer model includes simulated card images that are generated for the card type. The simulated card images are generated based on a semi-randomized background that varies in appearance, onto which an identifying marking of the card type is superimposed, such that the training data for the computer model includes additional randomized sample card images and ensure the model is robust to further variations in subtypes.
1. A system, comprising:
a processor that executes instructions;
a non-transitory computer-readable medium having instructions executable by the processor causing the processor to:
generate a set of simulated card images by adding an identifying marking of a card type to a set of semi-randomized card image backgrounds, the set of simulated card images simulating unknown subtypes of the card type; and
include the set of simulated card images in a training set of images for the card type, the training set for training a computer model to identify the card type with a card image based on the training set of images.
2. The system of claim 1 , wherein the instructions further cause the processor to provide the computer model to a system configured to train the computer model using the training set of images for the card type.
3. The system of claim 1 , wherein the instructions further cause the processor to generate the set of semi-randomized card image backgrounds.
4. The system of claim 1 , wherein the instructions further cause the processor to identify the identifying marking of the card type.
5. The system of claim 4 , wherein the identifying markings of the card type are determined by comparing features across known subtypes of the card type.
6. The system of claim 5 , wherein the features compared across known subtypes include keypoints or keypoint descriptors using an inlier-based analysis to determine a maximum number of matching keypoints.
7. The system of claim 1 , wherein the instructions to generate the set of simulated card images further comprises applying one or more distortions to the training set of images.
8. The system of claim 7 , wherein the distortions are based on the consistency of image capture expected for card images to be classified by the computer model.
9. The system of claim 1 , wherein the instructions to generate the set of simulated card images further comprises applying a transform to the identifying marking before adding the identifying marking to at least one of the card image backgrounds.
10. A method, comprising:
generating a set of simulated card images by adding an identifying marking of a card type to a set of semi-randomized card image backgrounds, the set of simulated card images simulating unknown subtypes of the card type; and
including the set of simulated card images in a training set of images for the card type, the training set for training a computer model to identify the card type with a card image based on the training set of images.
11. The method of claim 10 , further comprising providing the computer model to a system configured to train the computer model using the training set of images for the card type.
12. The method of claim 10 , further comprising generating the set of semi-randomized card image backgrounds.
13. The method of claim 10 , further comprising identifying the identifying marking of the card type.
14. The method of claim 13 , wherein the identifying markings of the card type are determined by comparing features across known subtypes of the card type.
15. The method of claim 14 , wherein the features compared across known subtypes include keypoints or keypoint descriptors using an inlier-based analysis to determine a maximum number of matching keypoints.
16. The method of claim 10 , generating the set of simulated card images further comprises applying one or more distortions to the training set of images.
17. The method of claim 7 , wherein the distortions are based on the consistency of image capture expected for card images to be classified by the computer model.
18. The method of claim 10 , generating the set of simulated card images further comprises applying a transform to the identifying marking before superimposing the identifying marking on at least one of the card image backgrounds.
19. A non-transitory computer-readable medium comprising instructions executable by a processor that cause the processor to:
generate a set of simulated card images by adding an identifying marking of a card type to a set of semi-randomized card image backgrounds, the set of simulated card images simulating unknown subtypes of the card type; and
include the set of simulated card images in a training set of images for the card type, the training set for training a computer model to identify the card type with a card image based on the training set of images.
20. The non-transitory computer-readable medium of claim 19 wherein the instructions further cause the processor to provide the computer model to a system configured to train the computer model using the training set of images for the card type.