Google to Improve Mobile Machine Learning Models

The team used multi-criteria optimization to achieve high speed and accuracy in the neural network MnasNet
09 August 2018   641

The neural network MnasNet uses a training method with reinforcement to select the appropriate mobile device architecture.

The MnasNet system contains:

  • object management based on a recurrent neural network for learning and selecting the architecture of the model;
  • he coach who builds the models and teaches their accuracy;
  • an output machine that calculates the speed of the model on the phones using TensorFlow.

The team used multi-criteria optimization to achieve high speed and accuracy. In addition, scientists used the learning algorithm with reinforcement and remuneration function. Thus, MnasNet finds Pareto optimality for each platform. For each mobile device, individual features are created that require a specific architecture.

Automated Neural Design Approach
Automated Neural Design Approach

To achieve the optimal balance between the flexibility of search and the scope of feasible solutions, a hierarchical approach has been applied. It represents a convolutional neural network in the form of a series of blocks. After, sequential search to assign layers of architecture to each block is being used. Due to this, each layer uses different operations and connections. In this case, the layers of blocks are copied by force.

MnasNet blocks
MnasNet blocks

Scientists have included the search speed of architecture in the function of rewarding the search algorithm.

MnasNet Blocks
MnasNet Blocks

According to the developers, the system picks up the model 1.5 times faster than MobileNetV2 and 2.4 times faster than NASNet, while taking into account the optimal accuracy and speed. As a proof, the result was demonstrated by the example of working with ImageNet image database.

Neural Network Now Can Animate People on Photos

Algorithm can even make people on the photos to 'go out' picture's borders
12 December 2018   110

Researchers at the University of Washington, together with the developers of Facebook, have created an algorithm that “revives” people in the photographs. In a single snapshot, it generates a three-dimensional moving model of a figure that can sit, jump, run, and even "go" beyond the limits of the image. The algorithm also works for drawings and anime characters.

To create such a technology, researchers used the experience of colleagues.

  • Mask R-CNN recognizes a human figure in the image and makes it stand out from the background.
  • Another algorithm imposes a simplified skeleton markup on the shape, defining how it will move.
  • The third algorithm "fills" the background space, previously hidden by the figure.

Further, the own algorithm of researchers on the basis of a marked two-dimensional figure creates a three-dimensional model and generates a texture level from the original image.

The developers added a user interface that allows you to change the shape of the figure in order to edit the photo itself or determine where the animation will begin. In addition, you can “revive” a drawing or photo in augmented reality and see a three-dimensional figure in VR or AR glasses.