Google Inc.

Menglong Zhu

Dmitry Kalenichenko

Andrew Gerald Howard

Mason Liu

Bo Chen

Mark M Sandler

Yinxiao Li

Liang-Chieh Chen

Andrey Zhmoginov

Marie Charisse White

Marco Andreetto

Weijun Wang

Tobias Christoph Weyand

Systems and methods for detecting objects in a video are provided. A method can include inputting a video comprising a plurality of frames into an interleaved object detection model comprising a plurality of feature extractor networks and a shared memory layer. For each of one or more frames, the operations can include selecting one of the plurality of feature extractor networks to analyze the one or more frames, analyzing the one or more frames by the selected feature extractor network to determine one or more features of the one or more frames, determining an updated set of features based at least in part on the one or more features and one or more previously extracted features extracted from a previous frame stored in the shared memory layer, and detecting an object in the one or more frames based at least in part on the updated set of features.

Memory-guided video object detection

The present disclosure provides directed to new, more efficient neural network architectures. As one example, in some implementations, the neural network architectures of the present disclosure can include a linear bottleneck layer positioned structurally prior to and/or after one or more convolutional layers, such as, for example, one or more depthwise separable convolutional layers. As another example, in some implementations, the neural network architectures of the present disclosure can include one or more inverted residual blocks where the input and output of the inverted residual block are thin bottleneck layers, while an intermediate layer is an expanded representation. For example, the expanded representation can include one or more convolutional layers, such as, for example, one or more depthwise separable convolutional layers. A residual shortcut connection can exist between the thin bottleneck layers that play a role of an input and output of the inverted residual block.

Highly efficient convolutional neural networks

The present disclosure provides systems and methods to reduce computational costs associated with convolutional neural networks. In addition, the present disclosure provides a class of efficient models termed 'MobileNets' for mobile and embedded vision applications. MobileNets are based on a straight-forward architecture that uses depthwise separable convolutions to build light weight deep neural networks. The present disclosure further provides two global hyper-parameters that efficiently trade-off between latency and accuracy. These hyper-parameters allow the entity building the model to select the appropriately sized model for the particular application based on the constraints of the problem. MobileNets and associated computational cost reduction techniques are effective across a wide range of applications and use cases.

Efficient convolutional neural networks and techniques to reduce associated computational costs

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing object detection. In one aspect, a method includes receiving multiple video frames. The video frames are sequentially processed using an object detection neural network to generate an object detection output for each video frame. The object detection neural network includes a convolutional neural network layer and a recurrent neural network layer. For each video frame after an initial video frame, processing the video frame using the object detection neural network includes generating a spatial feature map for the video frame using the convolutional neural network layer and generating a spatio-temporal feature map for the video frame using the recurrent neural network layer.

Object detection using spatio-temporal feature maps

Growing community of inventors

Los Angeles, CA, United States of America

Menglong Zhu