Abstract: Logic may reduce the size of runtime memory for deep neural network inference computations. Logic may determine, for two or more stages of a neural network, a count of shared block allocations, or shared memory block allocations, that concurrently exist during execution of the two or more stages. Logic may compare counts of the shared block allocations to determine a maximum count of the counts. Logic may reduce inference computation time for deep neural network inference computations. Logic may determine a size for each of the shared block allocations of the count of shared memory block allocations, to accommodate data to store in a shared memory during execution of the two or more stages of the cascaded neural network. Logic may determine a batch size per stage of the two or more stages of a cascaded neural network based on a lack interdependencies between input data.

Abstract: Systems and methods are provided that tune a convolutional neural network (CNN) to increase both its accuracy and computational efficiency. In some examples, a computing device storing the CNN includes a CNN tuner that is a hardware and/or software component that is configured to execute a tuning process on the CNN. When executing according to this configuration, the CNN tuner iteratively processes the CNN layer by layer to compress and prune selected layers. In so doing, the CNN tuner identifies and removes links and neurons that are superfluous or detrimental to the accuracy of the CNN.

Abstract: Systems and methods are provided that tune a convolutional neural network (CNN) to increase both its accuracy and computational efficiency. In some examples, a computing device storing the CNN includes a CNN tuner that is a hardware and/or software component that is configured to execute a tuning process on the CNN. When executing according to this configuration, the CNN tuner iteratively processes the CNN layer by layer to compress and prune selected layers. In so doing, the CNN tuner identifies and removes links and neurons that are superfluous or detrimental to the accuracy of the CNN.

Abstract: Logic may reduce the size of runtime memory for deep neural network inference computations. Logic may determine, for two or more stages of a neural network, a count of shared block allocations, or shared memory block allocations, that concurrently exist during execution of the two or more stages. Logic may compare counts of the shared block allocations to determine a maximum count of the counts. Logic may reduce inference computation time for deep neural network inference computations. Logic may determine a size for each of the shared block allocations of the count of shared memory block allocations, to accommodate data to store in a shared memory during execution of the two or more stages of the cascaded neural network. Logic may determine a batch size per stage of the two or more stages of a cascaded neural network based on a lack interdependencies between input data.

Abstract: A method for graphical hairstyle generation presents an interactive technique that produces static hairstyles by generating individual hair strands of the desired shape and color, subject to the presence of gravity and collisions. A variety of hairstyles can be generated by adjusting the wisp parameters while the deformation is solved efficiently accounting for the effects of gravity and collisions. Wisps are generated employing statistical approaches. As for hair deformation, a method is used based on physical simulation concepts but is simplified to efficiently solve the static shape of hair. On top of the statistical wisp model and the deformation solver, a constraint-based styler models artificial features that oppose the natural flow of hair under gravity and hair elasticity, such as a hairpin. Our technique spans a wider range of human hairstyles than previously proposed methods, and the styles generated by this technique are pretty realistic.

Abstract: A method for graphical hairstyle generation presents an interactive technique that produces static hairstyles by generating individual hair strands of the desired shape and color, subject to the presence of gravity and collisions. A variety of hairstyles can be generated by adjusting the wisp parameters while the deformation is solved efficiently accounting for the effects of gravity and collisions. Wisps are generated employing statistical approaches. As for hair deformation, a method is used based on physical simulation concepts but is simplified to efficiently solve the static shape of hair. On top of the statistical wisp model and the deformation solver, a constraint-based styler models artificial features that oppose the natural flow of hair under gravity and hair elasticity, such as a hairpin. Our technique spans a wider range of human hairstyles than previously proposed methods, and the styles generated by this technique are pretty realistic.