Fixation generation for machine learning
The disclosure extends to methods, systems, and apparatuses for automated fixation generation and more particularly relates to generation of synthetic saliency maps. A method for generating saliency information includes receiving a first image and an indication of one or more sub-regions within the first image corresponding to one or more objects of interest. The method includes generating and storing a label image by creating an intermediate image having one or more random points. The random points have a first color in regions corresponding to the sub-regions and a remainder of the intermediate image having a second color. Generating and storing the label image further includes applying a Gaussian blur to the intermediate image.
1. A method for generating an artificial saliency map, the method comprising:
receiving a first image and an indication of a ground truth bounding box within the first image corresponding to an object of interest;
generating an intermediate image that comprises random points within a region corresponding to the ground truth bounding box of the first image, wherein the intermediate image is a binary image such that it comprises only a first color and a second color;
applying a Gaussian blur to the intermediate image to generate a blurred intermediate image by applying a blur to each of the random points in the intermediate image;
downsizing the blurred intermediate image to generate a low resolution version of the blurred intermediate image; and
storing the low resolution version of the blurred intermediate image as a label image for the first image, wherein the first image and the label image are used for training or testing a neural network to determine information about any object of interest in any image.
2. The method of claim 1 , further comprising pairing the first image and the label image and feeding the pair of images to a modified deep neural network configured to output a synthetic saliency map to predict a location of an object within any image.
3. The method of claim 1 , wherein the blur applied to each of the random points in the intermediate image comprises an ellipses shaped blur configured to predict a scale and location of the object of interest.
4. The method of claim 1 , wherein the neural network is configured to generate a synthetic saliency map based on the first image and the label image.
5. The method of claim 4 , wherein the neural network consumes the first image and the label image and outputs the synthetic saliency map based on the first image and the label image.
6. The method of claim 5 , wherein the synthetic saliency map mimics human perception for object detection.
7. The method of claim 1 , further comprising determining how many random points to generate within the ground truth bounding box based on a size of the ground truth bounding box.
8. A processor that is programmable to execute instructions stored in non-transitory computer readable storage media, the instructions comprising:
receiving a first image and an indication of a ground truth bounding box within the first image corresponding to an object of interest;
generating an intermediate image that comprises random points within a region corresponding to the ground truth bounding box of the first image, wherein the intermediate image is a binary image such that it comprises only a first color and a second color;
applying a Gaussian blur to the intermediate image to generate a blurred intermediate image by applying a blur to each of the random points in the intermediate image;
downsizing the blurred intermediate image to generate a low resolution version of the blurred intermediate image; and
storing the low resolution version of the blurred intermediate image as a label image for the first image, wherein the first image and the label image are used for training a machine learning algorithm or model to output a synthetic saliency map or predict locations of objects of interest in images.
9. The processor of claim 8 , wherein the instructions further comprise pairing the first image and the label image and feeding the pair of images to a modified deep neural network configured to output the synthetic saliency map to predict a location of an object within any image.
10. The processor of claim 9 , wherein the instructions further comprise receiving ground truth about the object of interest within the ground truth bounding box, wherein the ground truth comprises one or more of a classification, an orientation, and a relative location of the object of interest.
11. The processor of claim 8 , wherein the machine learning algorithm or model is configured to consume the first image and the label image and output the synthetic saliency map, wherein the synthetic saliency map mimics human perception for object detection.
12. The processor of claim 8 , wherein the blur applied to each of the random points in the intermediate image comprises an ellipses shaped blur configured to predict a scale and location of the object of interest.
13. The processor of claim 8 , wherein the instructions further comprise determining how many random points to generate within the ground truth bounding box based on a size of the ground truth bounding box.
14. Non-transitory computer readable storage media storing instructions that, when executed by one or more processors, cause the one or more processors to:
receive a first image and an indication of a ground truth bounding box within the first image corresponding to an object of interest;
generate an intermediate image that comprises random points within a region corresponding to the ground truth bounding box of the first image, wherein the intermediate image is a binary image such that it comprises only a first color and a second color;
apply a Gaussian blur to the intermediate image to generate a blurred intermediate image by applying a blur to each of the random points in the intermediate image;
downsize the blurred intermediate image to generate a low resolution version of the blurred intermediate image; and
store the low resolution version of the blurred intermediate image as a label image for the first image, wherein the first image and the label image are used for training or testing a neural network to determine information about the object of interest.
15. The non-transitory computer readable storage media of claim 14 , wherein the instructions further cause the one or more processors to pair the first image and the label image and feed the pair of images to a modified deep neural network configured to output a synthetic saliency map to predict a location of an object within any image.
16. The non-transitory computer readable storage media of claim 14 , wherein the blur applied to each of the random points in the intermediate image comprises an ellipses shaped blur configured to predict a scale and location of the object of interest.
17. The non-transitory computer readable storage media of claim 14 , wherein the neural network is configured to generate a synthetic saliency map based on the first image and the label image, wherein the synthetic saliency map mimics human perception for object detection.
18. The non-transitory computer readable storage media of claim 14 , wherein the instructions further cause the one or more processors to determine how many random points to generate within the ground truth bounding box based on a size of the ground truth bounding box.