Methods and apparatus for high-fidelity vision tasks using deep neural networks

1. An apparatus comprising:

interface circuitry:

machine-readable instructions; and

at least one processor circuit to be programmed by the machine-readable instructions to:

extract low-level features and edge-enhanced features of an input image processed using a convolutional neural network;

generate an eidetic memory block using the extracted low-level features or the extracted edge-enhanced features;

perform image segmentation using the eidetic memory block, the eidetic memory block used to propagate domain-persistent features through a segmentation network, wherein the image segmentation includes skipping generation of inception features or atrous features to proceed to generation of collation features, the collation features generated based on the domain-persistent features of the eidetic memory block; and

apply the image segmentation to an object of interest in the input image to identify a pixel-level segmentation of the object.

2. The apparatus of claim 1 , wherein one or more of the at least one processor circuit is to cause the segmentation network to generate the inception features, the atrous features, or the collation features.

3. The apparatus of claim 1 , wherein the generation of the collation features assimilates multi-scale information from the inception features, the atrous features, or the eidetic memory block.

4. The apparatus of claim 1 , wherein the low-level features are extracted using a Visual Geometry Group (VGG) network.

5. The apparatus of claim 1 , wherein the edge-enhanced features are extracted using a wavelet transformation.

6. The apparatus of claim 1 , wherein the input image includes user-based input image feedback, the input image feedback including a positive click or a negative click to identify the object of interest within the input image.

7. A method, comprising:

extracting low-level features and edge-enhanced features of an input image processed using a convolutional neural network;

generating an eidetic memory block using the extracted low-level features or the extracted edge-enhanced features;

performing image segmentation using the eidetic memory block, the eidetic memory block used to propagate domain-persistent features through a segmentation network, wherein the image segmentation includes skipping generation of inception features or atrous features to proceed to generation of collation features, the collation features generated based on the domain-persistent features of the eidetic memory block; and

applying, by at least one processor circuit programmed by at least one instruction, the image segmentation to an object of interest in the input image to identify a pixel-level segmentation of the object.

8. The method of claim 7 , further including assimilating multi-scale information from the inception features, the atrous features, or the eidetic memory block.

9. The method of claim 7 , wherein the low-level features are extracted using a Visual Geometry Group (VGG) network.

10. The method of claim 7 , wherein the edge-enhanced features are extracted using a wavelet transformation.

11. The method of claim 7 , further including receiving user-based input image feedback, the input image feedback including a positive click or a negative click to identify the object of interest within the input image.

12. The method of claim 7 , wherein the performing of the image segmentation includes generating the inception features or generating the atrous features.

13. At least one non-transitory computer readable medium comprising instructions to cause at least one processor circuit to at least:

extract low-level features and edge-enhanced features of an input image processed using a convolutional neural network;

generate an eidetic memory block using the extracted low-level features or the extracted edge-enhanced features;

perform image segmentation using the eidetic memory block, the eidetic memory block used to propagate domain-persistent features through a segmentation network, wherein the image segmentation includes skipping generation of inception features or atrous features to proceed to generation of collation features, the collation features generated based on the domain-persistent features of the eidetic memory block; and

apply the image segmentation to an object of interest in the input image to identify a pixel-level segmentation of the object.

14. The at least one non-transitory computer readable medium as defined in claim 13 , wherein the instructions are to cause one or more of the at least one processor circuit to assimilate multi-scale information from the inception features, the atrous features, or the eidetic memory block.

15. The at least one non-transitory computer readable medium as defined in claim 13 , wherein the instructions are to cause one or more of the at least one processor circuit to extract the low-level features using a Visual Geometry Group (VGG) network.

16. The at least one non-transitory computer readable medium as defined in claim 13 , wherein the instructions are to cause one or more of the at least one processor circuit to extract the edge-enhanced features using a wavelet transformation.

17. The at least one non-transitory computer readable medium as defined in claim 13 , wherein the instructions are to cause one or more of the at least one processor circuit to receive user-based input image feedback, the input image feedback including a positive click or a negative click to identify the object of interest within the input image.

18. The at least one non-transitory computer readable medium as defined in claim 13 , wherein the instructions are to cause one or more of the at least one processor circuit to perform the image segmentation to generate the inception features or generate the atrous features.