Canva Pty Ltd

Described herein is a computer implemented method for automatically editing a digital design in a design application. The method includes receiving a natural language request to edit the digital design; and accessing a native format representation of the digital design. The native format representation includes one or more native design elements. The method further includes generating a descriptive format representation of the digital design by converting at least one of the one or more native design elements into a descriptive design element; generating at least one descriptive edit action based on the request and the descriptive format representation of the digital design; converting the at least one descriptive edit action into a native edit action; and implementing the native edit action, in the design application, to update the digital design.

Described herein is a computer implemented method. The method includes receiving, via an input device, first user input drawing an input shape and generating, based on the first user input, original drawing data that includes an ordered set of points that define the input shape. The original drawing is processed to generate an input vector which also includes an ordered set of points. The input shape is then classified as a first template shape by processing the input vector using a machine learning model. A new shape is then generated based on the first template shape and the original drawing data.

Embodiments relate to methods, systems and computer-readable media for editing images. An embodiment of a method for editing an image includes accessing a source image for editing, accessing a reference image, generating a mapping, wherein the mapping relates at least one source attribute value to at least one target attribute value, and modifying the source image to generate an edited image by applying the generated mapping, the edited image having at least one attribute value based on the attribute values of the reference image. Embodiments also relate to methods, systems and computer-readable media for training machine learning models. An embodiment of a method for training a machine learning model to edit an image, includes accessing a source image for editing, accessing a reference image, generating a source histogram for the source image and a reference histogram for the reference image, inputting the source histogram and the reference histogram into an encoder to generate a source histogram vector and a target histogram vector, concatenating the source histogram vector and the reference histogram vector into a concatenated vector, inputting the concatenated vector into the machine learning model, generate a mapping using the machine learning model, applying the mapping to the source image, calculating at least one loss function, and updating the parameters of the machine learning model on the basis of the at least one loss function.

A computer-implemented method for generating a prompt for a generative machine learning model is described. A user text prompt is encoded by an encoder implementing an encoding process, into a text embedding. The encoded text prompt is used to identify text embeddings in a vector database. The text embeddings in the vector database correspond to training captions used to train the generative machine learning model encoded using the encoding process. The identified text embeddings are used to generate a modified prompt. A large language model may be used to generate the modified prompt based on the identified text embeddings. The modified prompt may be passed to a generative machine learning model to generate media or one or more digital assets, such as a text-to-image machine learning model to generate one or more images.

Described herein is a computer implemented method. The method includes accessing a first set of source objects that relate to a first page of a source design, wherein each source object is associated with a set of source object attributes that include a source identifier attribute that identifies source content of the source design; a type attribute that identifies a content type of the source content; and a reading order attribute that identifies a position of the source content in a reading order of the source design. The method further includes accessing a first set of destination objects that relate to a first page of a destination design and generating, based on the first set of source objects and the first set of destination objects, a first content map that maps each source object in the first set of source objects to a destination object in the first set of destination objects.

Described herein is a computer implemented method. The method includes receiving, via an input device, first user input drawing an input shape and generating, based on the first user input, original drawing data that includes an ordered set of points that define the input shape. The original drawing is processed to generate an input vector which also includes an ordered set of points. The input shape is then classified as a first template shape by processing the input vector using a machine learning model. A new shape is then generated based on the first template shape and the original drawing data.

Described herein is a computer implemented method for generating an alert based on a product update for a software system, the product update including a first product update variant of the software system and a product update control of the software system, the method including: receiving user interaction event data based on one or more user interaction events from a plurality of users of the software system, the plurality of users including a first subset of users using the first product update variant and a second subset of users using the product update control; processing the user interaction event data to determine if any of the one or more user interaction events is an exposure event; based on the user interaction event data of any determined exposure events, determining, for each of the first product update variant and product update control, a respective user metric dataset for a user metric; calculating, for the first product update variant and product update control, respective probability input data in respect of the user metric based on the respective user metric datasets; based on the probability input data, calculating, for the first product update variant and product update control, a first variant estimation model and a control estimation model respectively; deriving a first change value distribution based on the first variant estimation model and the control estimation model such that the first change value distribution defines probability of relative uplift or downlift of user interactions; calculating a first impact value based on first change value distribution and a predefined time value t; comparing the calculated first impact value with a predefined impact value threshold; and based on this comparison, generating a first alert condition.

Computer processing systems and computer implemented methods for determining an arrangement of elements of a design are described. The arrangement may be into rows or columns. Collections of elements for rows and/or columns are determined based on overlap of the elements along a dimension prior to arrangement. Each collection may be arranged within an area of the design into a row or column. In some embodiments an initial arrangement into both rows and columns is determined and a final arrangement is determined based on which arrangement results in a least total overlap area of the elements.