System improvement for deep neural networks
The technology relates to tuning a data translation block (DTB) including a generator model and a discriminator model. One or more processors may be configured to receive training data including an image in a second domain. The image in the second domain may be transformed into a first domain with a generator model. The transformed image may be processed to determine one or more outputs with one or more deep neural networks (DNNs) trained to process data in the first domain. An original objective function for the DTB may be updated based on the one or more outputs. The generator and discriminator models may be trained to satisfy the updated objective function.
1. A method for processing image data, the method including:
receiving, by one or more processors, an image in a second domain;
transforming, by an optimized data translation block (ODTB), the image from the second domain to a transformed image in a first domain;
processing, the transformed image to determine one or more outputs.
2. The method of claim 1 , wherein the trained model includes one or more DNNs trained to process data in the first domain.
3. The method of claim 2 , wherein each of the one or more DNNs outputs a respective output of the one or more outputs, and
each respective output includes one or more predictions or estimated values.
4. The method of claim 2 , wherein the ODTB is trained using training data including a set of images in the second domain.
5. The method of claim 4 , wherein the training data further includes, for each image in the set of images, markings, the markings being separated into subsets, with each subset of markings being assigned to the one or more DNNs.
6. The method of claim 5 , further comprising training the ODTB, wherein the training includes:
updating an original objective function for the ODTB based on the one or more outputs.
7. The method of claim 5 , wherein the training includes calculating a loss value (Loss i ) for each of the one or more outputs, the loss value being calculated using the formula:
Loss i =(M D2−i −M ˜D2−i ) 2 , where i is the DNN index, where M D2−i is the subset of markings assigned to DNNi and M ˜D2−i is the one or more predictions or estimated values output by DNNi.
8. The method of claim 7 , wherein the training further includes:
weighting the loss value for each of the one or more DNNs;
aggregating the weighted loss value for each of the one or more DNNs; and
determining an average loss value (DNN_AVG_LOSS).
9. The method of claim 8 , wherein the updated objective function is:
O CGAN +α DNN_AVG_LOSS *DNN_AVG_LOSS, where O CGAN is the original objective function of the DTB and α DNN_AVG_LOSS is a weight factor.
10. The method of claim 4 , wherein the training data further includes a training image in the first domain.
11. A system including:
one or more computing devices having one or more processors; and
memory storing instructions, the instructions, when executed by the one or more processors, cause the one or more processors to:
receive an image in a second domain;
transform, by an optimized data translation block (ODTB), the image from the second domain to a transformed image in a first domain; and
processing, the transformed image by a trained model to determine one or more outputs.
12. The system of claim 11 , wherein the trained model includes one or more DNNs trained to process data in the first domain.
13. The system of claim 12 , wherein each of the one or more DNNs outputs a respective output of the one or more outputs, and
each respective output includes one or more predictions or estimated values.
14. The system of claim 12 , wherein the ODTB is trained using training data including a set of images in the second domain.
15. The system of claim 14 , wherein the training data further includes, for each image in the set of images, markings, the markings being separated into subsets, with each subset of markings being assigned to the one or more DNNs.
16. The system of claim 15 , wherein the instructions, when executed, further cause the one or more processors to train the ODTB, wherein the training includes:
updating an original objective function for the ODTB based on the one or more outputs.
17. The system of claim 15 , wherein the training includes calculating a loss value (Loss i ) for each of the one or more outputs, the loss value being calculated using the formula:
Loss i =(M D2−i −M ˜D2−i ) 2 , where i is the DNN index, where M D2−i is the subset of markings assigned to DNNi and M ˜D2−i is the one or more predictions or estimated values output by DNNi.
18. The system of claim 17 , wherein the training further includes:
weighting the loss value for each of the one or more DNNs;
aggregating the weighted loss value for each of the one or more DNNs; and
determining an average loss value (DNN_AVG_LOSS).
19. The system of claim 18 , wherein the updated objective function is:
O CGAN +α DNN_AVG_LOSS *DNN_AVG_LOSS, where O CGAN is the original objective function of the DTB and α DNN_AVG_LOSS is a weight factor.
20. The system of claim 14 , wherein the training data further includes a training image in the first domain.