Google Inc.

Neil Matthew Tinmouth Houlsby

Alexey Dosovitskiy

Dirk Weissenborn

Alexander Kolesnikov

Lucas Klaus Beyer

Xiaohua Zhai

Matthias Johannes Lorenz Minderer

Sylvain Gelly

Jakob D Uszkoreit

Thomas Unterthiner

Mostafa Dehghani

Georg Heigold

Alexey Alexeevich Gritsenko

Austin Charles Stone

Yin Ching Jessica Yung

Basil Mustafa

Joan Puigcerver I Perez

Daniel M Keysers

Vivek Natarajan

Mario Lucic

Michael Tobias Tschannen

Ilya Tolstikhin

Wen Yau Aaron Loh

Patricia Leigh MacWilliams

André Susano Pinto

Jan Freyberg

Alan Prasana Karthikesalingam

Carlos Riquelme Ruiz

Andreas Peter Steiner

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training neural networks using transfer learning. One of the methods includes training a neural network to perform a first prediction task, including: obtaining trained model parameters for each of a plurality of candidate neural networks, wherein each candidate neural network has been pre-trained to perform a respective second prediction task that is different from the first prediction task; obtaining a plurality of training examples corresponding to the first prediction task; selecting a proper subset of the plurality of candidate neural networks using the plurality of training examples; generating, for each candidate neural network, one or more fine-tuned neural networks, wherein each fine-tuned neural network is generated by updating the model parameters of the candidate neural network using the plurality of training examples; and determining model parameters for the neural network using the respective fine-tuned neural networks.

Training neural networks using transfer learning

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing images using mixer neural networks. One of the methods includes obtaining one or more images comprising a plurality of pixels; determining, for each image of the one or more images, a plurality of image patches of the image, wherein each image patch comprises a different subset of the pixels of the image; processing, for each image of the one or more images, the corresponding plurality of image patches to generate an input sequence comprising a respective input element at each of a plurality of input positions, wherein a plurality of the input elements correspond to respective different image patches; and processing the input sequences using a neural network to generate a network output that characterizes the one or more images, wherein the neural network comprises one or more mixer neural network layers.

Multi-layer perceptron-based computer vision neural networks

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network to perform a downstream computer vision task. One of the methods includes pre-training an initial neural network that shares layers with the neural network to perform an initial computer vision task and then training the neural network on the downstream computer vision task.

Transfer learning between different computer vision tasks

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for object detection. In one aspect, a method comprises: obtaining: (i) an image, and (ii) a set of one or more query embeddings, wherein each query embedding represents a respective category of object; processing the image and the set of query embeddings using an object detection neural network to generate object detection data for the image, comprising: processing the image using an image encoding subnetwork of the object detection neural network to generate a set of object embeddings; processing each object embedding using a localization subnetwork to generate localization data defining a corresponding region of the image; and processing: (i) the set of object embeddings, and (ii) the set of query embeddings, using a classification subnetwork to generate, for each object embedding, a respective classification score distribution over the set of query embeddings.

Open-vocabulary object detection in images

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing images using self-attention based neural networks. One of the methods includes obtaining one or more images comprising a plurality of pixels; determining, for each image of the one or more images, a plurality of image patches of the image, wherein each image patch comprises a different subset of the pixels of the image; processing, for each image of the one or more images, the corresponding plurality of image patches to generate an input sequence comprising a respective input element at each of a plurality of input positions, wherein a plurality of the input elements correspond to respective different image patches; and processing the input sequences using a neural network to generate a network output that characterizes the one or more images, wherein the neural network comprises one or more self-attention neural network layers.

Processing images using self-attention based neural networks

Growing community of inventors

Zurich, Switzerland

Neil Matthew Tinmouth Houlsby