Snap Inc.

Maksim Gusarov

Igor Kudriashov

Fedir Poliakov

Sergei Kotcur

Sergey Tulyakov

Aleksei Stoliar

Sergei Korolev

Christopher Yale Crutchfield

Andrew Wan

Valerii Filev

Chen Wang

Patrick Poirson

Zhenpeng Zhou

Oleg Tovstyi

An input video item that includes a target visual augmentation is accessed. A machine learning model uses the input video item to generate an embedding. The embedding may comprise a vector representation of a visual effect of the target visual augmentation. The machine learning model is trained, in an unsupervised training phase, to minimize loss between training video representations generated within each of a plurality of training sets. Each training set comprises a plurality of different training video items that each include a predefined visual augmentation. Based on the generation of the embedding of the input video item, the target visual augmentation is mapped to an augmentation identifier.

Embeddings representing visual augmentations

A compact generative neural network can be distilled from a teacher generative neural network using a training network. The compact network can be trained on the input data and output data of the teacher network. The training network train the student network using a discrimination layer and one or more types of losses, such as perception loss and adversarial loss.

Generative neural network distillation

Systems, devices, media, and methods are presented for segmenting an image of a video stream with a client device, identifying an area of interest, generating a modified area of interest within one or more image, identifying a first set of pixels and a second set of pixels, and modifying a color value for the first set of pixels.

Image segmentation and modification of a video stream

Systems, devices, media, and methods are presented for generating facial representations using image segmentation with a client device. The systems and methods receive an image depicting a face, detect at least a portion of the face within the image, and identify a set of facial landmarks within the portion of the face. The systems and methods determine one or more characteristics representing the portion of the face, in response to detecting the portion of the face. Based on the one or more characteristics and the set of facial landmarks, the systems and methods generate a representation of a face.

Automated avatar generation

Systems, devices, and methods are presented for segmenting an image of a video stream with a client device by accessing a set of images within a video stream, identifying an object of interest within one or more images of the set of images, and detecting a region of interest within the one or more images. The systems, devices, and method identify a first set of median pixels in a first portion of the object of interest and a second set of median pixels in a second portion of the object of interest. The systems, devices, and methods determine a polyline approximating the first and second sets of median pixels and generate a model for the polyline.

Image segmentation for object modeling

Growing community of inventors

Marina Del Rey, CA, United States of America

Maksim Gusarov