Abstract: Systems, methods, and non-transitory computer-readable media can identify an object depicted in an image. True color information associated with the object is obtained from a true color database comprising true color information for a plurality of objects. A color delta associated with the object is determined based on the true color information and captured color information associated with the object. The image is modified based on the color delta.

Abstract: The disclosure is directed to a push notification system (PNS) for sending notifications to users of an application (“app”), e.g., a mobile app of a social networking application. The PNS generates a compact filter data structure (“filter”) to store a representation of token identifications (“IDs”) of app installations that belong in a specified category. A token ID uniquely identifies a device-app pair. To send a notification, the PNS identifies the filter corresponding to the specified category, tests the filter with token IDs of all installations of the app to identify which of the token IDs are present in the filter, and sends the notification to client devices associated with the identified set of token IDs. The filter facilitates a fast lookup of a token ID while consuming significantly less storage space, e.g., as the filter stores a representation of the token IDs and not the actual token IDs.

Abstract: Systems, methods, and non-transitory computer-readable media can identify an object depicted in an image. True color information associated with the object is obtained from a true color database comprising true color information for a plurality of objects. A color delta associated with the object is determined based on the true color information and captured color information associated with the object. The image is modified based on the color delta.

Abstract: The disclosure is directed to a push notification system (PNS) for sending notifications to users of an application (“app”), e.g., a mobile app of a social networking application. The PNS generates a compact filter data structure (“filter”) to store a representation of token identifications (“IDs”) of app installations that belong in a specified category. A token ID uniquely identifies a device-app pair. To send a notification, the PNS identifies the filter corresponding to the specified category, tests the filter with token IDs of all installations of the app to identify which of the token IDs are present in the filter, and sends the notification to client devices associated with the identified set of token IDs. The filter facilitates a fast lookup of a token ID while consuming significantly less storage space, e.g., as the filter stores a representation of the token IDs and not the actual token IDs.

Abstract: Some embodiments include a computer server implementing a splitter engine to perform testing of a pre-production version of an application service against live traffic. The splitter engine can receive a client request for a live application service. The splitter engine can generate a live response by processing the client request through a live instance of a production engine. The splitter engine can select one or more request processors to process the client request and a processing order based on a request type of the client request. The splitter engine can process the client request through the one or more request processors according to the processing order after responding to the client request with the live response. The splitter engine can compare states of at least two of the production engine and the one or more request processors.

Abstract: Some embodiments include a computer server implementing a splitter engine to perform testing of a pre-production version of an application service against live traffic. The splitter engine can receive a client request for a live application service. The splitter engine can generate a live response by processing the client request through a live instance of a production engine. The splitter engine can select one or more request processors to process the client request and a processing order based on a request type of the client request. The splitter engine can process the client request through the one or more request processors according to the processing order after responding to the client request with the live response. The splitter engine can compare states of at least two of the production engine and the one or more request processors.