Randomly generated blobs to improve object-detection training for framed video content
Generating a training image for use in training a region-of-interest detector that is trained to detect regions-of-interest within images includes generating a closed geometric shape; filling the closed geometric shape with a filler to obtain a blob; overlaying the blob on an edge of an image to obtain the training image, where the image includes a region-of-interest and a background region, and where the edge separates the region-of-interest from the background region; and using the training image to train the region-of-interest detector to detect a boundary of the region-of-interest. An input to the region-of-interest detector in a training phase includes the training image and a first indication of coordinates of the region-of-interest in the training image. An output of the region-of-interest detector includes a second indication of an area of the training image and a probability of the area of the training image being the region-of-interest.
1. A method for generating a training image for use in training a region-of-interest detector, wherein the region-of-interest detector is trained to detect regions-of-interest within images, comprising:
generating a closed geometric shape;
filling the closed geometric shape with a filler to obtain a blob;
overlaying the blob on an edge of an image to obtain the training image,
wherein the image comprises a region-of-interest and a background region, and
wherein the edge separates the region-of-interest from the background region; and
using the training image to train the region-of-interest detector to detect a boundary of the region-of-interest,
wherein an input to the region-of-interest detector in a training phase comprises:
the training image; and
a first indication of coordinates of the region-of-interest in the training image; and
wherein an output of the region-of-interest detector comprises:
a second indication of an area of the training image, and
a probability of the area of the training image being the region-of-interest.
2. The method of claim 1 , wherein generating the closed geometric shape comprises:
obtaining the closed geometric shape using a Fourier series.
3. The method of claim 2 , wherein obtaining the closed geometric shape using the Fourier series comprises:
obtaining the closed geometric shape using a random number of sine waves or random amplitudes of Fourier coefficients.
4. The method of claim 1 , wherein the filler is a random noise.
5. The method of claim 1 , wherein the filler is a single color.
6. The method of claim 1 , wherein the filler is a filler image.
7. The method of claim 1 , wherein the blob is overlaid on a vertex of the region-of-interest.
8. An apparatus for generating a training image for use in training a region-of-interest detector, wherein the region-of-interest detector is trained to detect regions-of-interest within images, the apparatus comprising:
a processor configured to:
generate a closed geometric shape;
fill the closed geometric shape with a filler to obtain a blob;
overlay the blob on an edge of an image to obtain the training image,
wherein the image comprises a region-of-interest and a background region, and
wherein the edge separates the region-of-interest from the background region; and
train, using the training image, the region-of-interest detector to detect a boundary of the region-of-interest,
wherein an input to the region-of-interest detector in a training phase comprises the training image, and
wherein an output of the region-of-interest detector comprises an indication of an area of the training image as being the region-of-interest.
9. The apparatus of claim 8 , wherein to generate the closed geometric shape comprises to:
obtain the closed geometric shape using a Fourier series.
10. The apparatus of claim 9 , wherein to obtain the closed geometric shape using the Fourier series comprises to:
obtain the closed geometric shape using a random number of sine waves or random amplitudes of Fourier coefficients.
11. The apparatus of claim 8 , wherein the filler is a random noise.
12. The apparatus of claim 8 , wherein the filler is a single color.
13. The apparatus of claim 8 , wherein the filler is a filler image.
14. The apparatus of claim 8 , wherein the blob is overlaid on a vertex of the region-of-interest.
15. A non-transitory computer-readable storage medium, comprising executable instructions that, when executed by a processor, facilitate performance of operations for generating a training image for use in training a region-of-interest detector, wherein the region-of-interest detector is trained to detect regions-of-interest within images, the operations comprising operations to:
generate a closed geometric shape;
fill the closed geometric shape with a filler to obtain a blob; and
overlay the blob on an edge of an image to obtain the training image,
wherein the image comprises a region-of-interest and a background region, and
wherein the edge separates the region-of-interest from the background region.
16. The non-transitory computer-readable storage medium of claim 15 , wherein to generate the closed geometric shape comprises to:
obtain the closed geometric shape using a Fourier series.
17. The non-transitory computer-readable storage medium of claim 16 , wherein to obtain the closed geometric shape using the Fourier series comprises to:
obtain the closed geometric shape using a random number of sine waves or random amplitudes of Fourier coefficients.
18. The non-transitory computer-readable storage medium of claim 15 , wherein the filler is a random noise.
19. The non-transitory computer-readable storage medium of claim 15 , wherein the filler is a single color.
20. The non-transitory computer-readable storage medium of claim 15 , wherein the blob is overlaid on a vertex of the region-of-interest.