Classification and localization based on annotation information
Systems and techniques for classification and localization based on annotation information are presented. In one example, a system trains a convolutional neural network based on training data and a plurality of images. The training data is associated with a plurality of patients from at least one imaging device. The plurality of images is associated with a plurality of masks from a plurality of objects. The convolutional neural network comprises a decoder consisting of at least one up-sampling layer and at least one convolutional layer. The system also generates a loss function based on the plurality of masks, where the loss function is iteratively back propagated to tune parameters of the convolutional neural network. The system also predicts a classification label for an input image based on the convolutional neural network.
1. A machine learning system that employs masks and decoders for improved performance, comprising:
a processor that executes computer-executable instructions stored in a memory, which cause the processor to:
train a convolutional neural network based on a plurality of masks, the convolutional neural network comprising a set of decoders respectively consisting of at least one up-sampling layer and at least one convolutional layer, wherein a number of decoders in the set of decoders is determined during training of the convolutional neural network;
generate and back propagate a loss function based on the plurality of masks to tune the convolutional neural network;
classify an input image based on the convolutional neural network; and
generate a multi-dimensional visualization associated with the classified input image.
2. The machine learning system of claim 1 , wherein the convolutional neural network comprises a pretrained classifier network that outputs convolutional feature maps.
3. The machine learning system of claim 2 , wherein the convolutional neural network comprises a classification/localization network that outputs corresponding scoring maps based on the convolutional feature maps.
4. The machine learning system of claim 2 , wherein a size of a mask from the plurality of masks is matched with a size of a convolutional feature map from the convolutional feature maps.
5. The machine learning system of claim 2 , wherein a size of a mask from the plurality of masks is matched with a size of a convolutional feature map from the convolutional feature maps based on a max pooling process.
6. The machine learning system of claim 1 , wherein the processor employs the set of decoders to generate a localization map.
7. The machine learning system of claim 1 , wherein the processor generates the loss function based on a probability for a class associated with the plurality of masks.
8. A computer-implemented method that employs masks and decoders for improved performance, comprising: training, by a device operatively coupled to a processor, a convolutional neural network based on a plurality of masks, the convolutional neural network comprising a set of decoders respectively consisting of at least one up-sampling layer and at least one convolutional layer, wherein a number of decoders in the set of decoders is determined during training of the convolutional neural network; generating, by the device, and back propagating, by the device, a loss function based on the plurality of masks to tune the convolutional neural network; classifying, by the device, an input image based on the convolutional neural network; and generating, by the device, a multi-dimensional visualization associated with the classified input image.
9. The computer-implemented method of claim 8 , wherein the convolutional neural network comprises a pretrained classifier network that outputs convolutional feature maps.
10. The computer-implemented method of claim 9 , wherein the convolutional neural network comprises a classification/localization network that outputs corresponding scoring maps based on the convolutional feature maps.
11. The computer-implemented method of claim 9 , wherein a size of a mask from the plurality of masks is matched with a size of a convolutional feature map from the convolutional feature maps.
12. The computer-implemented method of claim 9 , wherein a size of a mask from the plurality of masks is matched with a size of a convolutional feature map from the convolutional feature maps based on a max pooling process.
13. The computer-implemented method of claim 8 , further comprising:
employing, by the device, the set of decoders to generate a localization map.
14. The computer-implemented method of claim 8 , wherein the generating the loss function is based on a probability for a class associated with the plurality of masks.
15. A computer program product for improving performance of a machine learning system via employment of masks and decoders, the computer program product comprising a computer readable memory having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to: train a convolutional neural network based on a plurality of masks, the convolutional neural network comprising a set of decoders respectively consisting of at least one up-sampling layer and at least one convolutional layer, wherein a number of decoders in the set of decoders is determined during training of the convolutional neural network; generate and back propagate a loss function based on the plurality of masks to tune the convolutional neural network; classify an input image based on the convolutional neural network; and generate a multi-dimensional visualization associated with the classified input image.
16. The computer program product of claim 15 , wherein the convolutional neural network comprises a pretrained classifier network that outputs convolutional feature maps.
17. The computer program product of claim 16 , wherein the convolutional neural network comprises a classification/localization network that outputs corresponding scoring maps based on the convolutional feature maps.
18. The computer program product of claim 16 , wherein a size of a mask from the plurality of masks is matched with a size of a convolutional feature map from the convolutional feature maps.
19. The computer program product of claim 16 , wherein a size of a mask from the plurality of masks is matched with a size of a convolutional feature map from the convolutional feature maps based on a max pooling process.
20. The computer program product of claim 15 , wherein the program instructions are further executable to cause the processor to:
employ the set of decoders to generate a localization map.