Abstract: In one embodiment, a method includes a system accessing an image, which may comprise covered and uncovered portions, and an overlay image comprising opaque pixels. The covered portion may be configured to be covered by the opaque pixels of the overlay image. The system may generate a data structure comprising data elements associated with pixels of the image. Each of the data elements associated with a covered pixel in the covered portion of the image may be configured to identify an uncovered pixel in the uncovered portion of the image that is closest to the covered pixel. Each covered pixel in the covered portion of the image may be modified by accessing the data element associated with the covered pixel, determining a distance between the covered pixel and an associated closest uncovered pixel using the accessed data element, and modifying a color of the covered pixel based on the distance.

Abstract: In one embodiment, a method for distributing an augmented reality (AR) effect may include ranking, by a computing system, a plurality of client devices based on a device capability, where a higher ranking may represent relatively better device capability. The system may identify a first subset of client devices having a first particular ranking. An AR effect may be transmitted to the first subset of client devices. The system may determine a performance score for the AR effect among the first subset of client devices. The performance score may be based on one or more performance parameters associated with the AR effect running on the first subset of client devices. The system may transmit the AR effect to a second subset of client devices when the performance score of the AR effect among the first subset of client devices satisfies a predefined threshold.

Abstract: A media effects engine on a computer device applies one or more effects to an input media stream. A performance monitor monitors a performance metric associated with playing the input media stream and reduces a quality parameter associated with the effect upon detecting a drop in the performance metric below a target metric. The quality parameter manages a tradeoff between a quality of effect and an amount of hardware resources consumed to produce the effect. Thus, the effect can be adjusted to meet the capabilities of the computer device.

Abstract: Systems, methods, and non-transitory computer-readable media can receive an image. One or more concepts depicted in the image are identified based on machine learning techniques. The one or more concepts are filtered based on filtering criteria to identify one or more selected concepts. An image description is generated comprising the one or more selected concepts.

Abstract: A video effects application executes on a client device having an image capture device and receives video data captured by the image capture device. The video effects application extracts information from the captured video data and stores the extracted information as metadata associated with the captured video data. For example, the video effects application identifies objects in the captured video data or identifies optical flow of the captured video data and stores the identified objects or identified optical flow as metadata associated with the captured video data. The video effects application stores information describing modifications to the captured video data in association with the captured video data. When the captured video data is presented, the captured video data, associated metadata, and information describing the modifications is communicated to a renderer, which uses the metadata to perform the identified modifications to the captured video data when presenting the captured video data.