Computer vision systems and methods for unsupervised learning for progressively aligning noisy contours
Computer vision systems and methods for noisy contour alignment are provided. The system generates a loss function and trains a convolutional neural network with the loss function and a plurality of images of a dataset to learn to align contours with progressively increasing complex forward and backward transforms over increasing scales. The system can align an attribute of an image of the dataset by the trained neural network.
1 . A computer vision system for performing noisy contour alignment comprising:
a memory; and
a processor in communication with the memory, the processor:
generating a loss function;
training a neural network to learn to align contours using the loss function, a cascade of warp predictors to predict transformations to align source features to target features, and a plurality of images; and
aligning an attribute of an image using the trained neural network;
wherein the cascade of warp predictors predicts transformations to align the source features to the target features at different scales; and
wherein each warp predictor concatenates the source features warped by an upsampled warp field, the target features, and the upsampled warp field.
2 . The system of claim 1 , wherein the processor determines an upper bound of the loss function.
3 . The system of claim 1 , wherein the loss function is a local shape-dependent Chamfer upper bound loss function that measures proximity and local shape similarity.
4 . The system of claim 1 , wherein the neural network is a convolutional neural network.
5 . The system of claim 1 , wherein the loss function is generated from a Modified National Institute of Standards and Technology (MNIST) digit image dataset or a geo-parcel to aerial image alignment dataset.
6 . The system of claim 1 , where the processor aligns a noisy digit contour present in the image or a parcel present in the image.
7 . A method for performing noisy contour alignment by a computer vision system, comprising the steps of:
generating a loss function;
training a neural network to learn to align contours using the loss function, a cascade of warp predictors to predict transformations to align source features to target features, and a plurality of images; and
aligning an attribute of an image using the trained neural network;
wherein the cascade of warp predictors predicts transformations to align the source features to the target features at different scales; and
wherein each warp predictor concatenates the source features warped by an upsampled warp field, the target features, and the upsampled warp field.
8 . The method of claim 7 , further comprising the step of determining an upper bound of the loss function.
9 . The method of claim 7 , wherein the loss function is a local shape-dependent Chamfer upper bound loss function that measures proximity and local shape similarity.
10 . The method of claim 8 , wherein the neural network is a convolutional neural network.
11 . The method of claim 7 , wherein the loss function is generated from a Modified National Institute of Standards and Technology (MNIST) digit image dataset or a geo-parcel to aerial image alignment dataset.
12 . The method of claim 7 , further comprising the step of aligning the attribute of the image by the trained neural network by aligning a noisy digit contour present in the image or a parcel present in the image.
13 . A non-transitory computer readable medium having instructions stored thereon for performing noisy contour alignment by a computer vision system which, when executed by a processor, causes the processor to carry out the steps of:
generating a loss function;
training a neural network to learn to align contours using the loss function, a cascade of warp predictors to predict transformations to align source features to target features, and a plurality of images; and
aligning an attribute of an image using the trained neural network;
wherein the cascade of warp predictors predicts transformations to align the source features to the target features at different scales; and
wherein each warp predictor concatenates the source features warped by an upsampled warp field, the target features, and the upsampled warp field.
14 . The non-transitory computer readable medium of claim 13 , the processor further carrying out the step of determining an upper bound of the loss function.
15 . The non-transitory computer readable medium of claim 13 , wherein the loss function is a local shape-dependent Chamfer upper bound loss function that measures proximity and local shape similarity.
16 . The non-transitory computer readable medium of claim 13 , the processor further carrying out the step of aligning the attribute of the image by the trained neural network by aligning a noisy digit contour present in the image or a parcel present in the image.