Dropbox, Inc.

Daniel R Horn

Ken Elkabany

Keith Winstein

Adhiraj Somani

Jongmin Baek

Nipunn Koorapati

Matteus Pan

Jean-Denis Greze

Elmer Charles Jubb, Iv

Bo Hu

Richard Chan

Philip Rha

Anatoly Yakovenko

Austin Sung

Andraz Kavalar

Andrew Sean Yang

Michael Loh

David Lichtenberg

Shi Feng

Andy Yang

Techniques for lossless compression of a digital image using prior image context.

Lossless decompression of digital images using prior image context

Lossless compression of digital images using prior image context

In some embodiments, an interface of a content management system manages synchronized content on storage systems. For example, the interface stores, on a metadata storage structure, records of metadata associated with blocks of data stored on a storage, the records including block identifiers that uniquely identify the blocks and timestamps associated with the blocks. The interface identifies a batch of storage operations associated with the blocks, including one or more delete operations. For each delete operation, the interface queries the metadata storage structure for a timestamp corresponding to a block of data associated with the delete operation, determines whether the delete operation creates a race condition between the delete operation and an add operation associated with the block of data, and rejects the delete operation when the delete operation creates the race condition or the timestamp corresponding to the block of data is newer than a predetermined period of time.

Storage interface for synchronizing content

Systems and methods for dynamic and automatic data storage scheme switching in a distributed data storage system. A machine learning-based policy for computing probable future content item access patterns based on historical content item access patterns is employed to dynamically and automatically switch the storage of content items (e.g., files, digital data, photos, text, audio, video, streaming content, cloud documents, etc.) between different data storage schemes. The different data storage schemes may have different data storage cost and different data access cost characteristics. For example, the different data storage schemes may encompass different types of data storage devices, different data compression schemes, and/or different data redundancy schemes.

Data storage scheme switching in a distributed data storage system

An interface of a content management system manages storage and access of content on the system. For example, after receiving, from a client, a request to download a content item, the interface determines whether the request includes a valid token. If so, the interface sends a content item request to a storage service, retrieves the content item, and sends the content item to the client. Otherwise, the interface sends an authorization request to an authorization service, an authentication request to an authentication service, and a content item request to the storage service. Based on the requests, the interface determines whether the content item is available in storage and whether the client is authorized to access the content item. When the content item is available in storage and the client is authorized to access the content item, the interface retrieves the content item and sends the content item to the client.

Techniques for image recompression. In one implementation, the techniques are implemented to recompress a baseline joint photographic experts group (JPEG) image with compression savings. The techniques replace the Huffman coding of baseline JPEG with an arithmetic coding that uses a sophisticated adaptive probability model. The arithmetic coding techniques avoid global operations such as global sorting that would prevent distributed and multithreaded decompression operations when recovering the original JPEG image from the recompressed image. At the same time, the techniques realize substantial compression savings relative to baseline JPEG, on average 23% compressing savings in some implementations.

Digital image recompression

A document preview system provides previews of a native document to client devices. To generate the preview of a document, the document preview system generates a preview document that presents the native document in a browser-compatible format including elements that correspond to fields in the native document. The document preview system also identifies a mapping between fields of the native document and the elements of the preview. The document preview system also identifies a formula listing for the elements of the preview code. When a user edits a document, the edits may be committed to the document preview system and in parallel generate a preview to update the user's display. When an updated preview is available, the updated preview is rendered in a hidden frame and the user's input context is captured and reproduced when the updated preview replaces the prior preview.

Modifying native documents with formulas in a preview

A document preview system provides previews of a native document to client devices. The previews include comments associated with native objects in the native document. The document preview system identifies bounding areas in the preview associated with the native objects, which may be identified by the rendering markers applied to the native document prior to rendering. Using the bounding areas, the document preview system identifies comments for the native document and determines the location to display the comment using native objects associated with the comment. When a new comment is received from a user for a preview of a native document, the document preview system determines native objects for the new comment that match a user's selection for placing the new comment. The new comment is inserted with the native objects in the native document.

Modifying native document comments in a preview

Lossless compression of a content item using a neural network trained on content item cohorts. A computing system includes a neural network that is used to train a plurality of symbol prediction models. Each symbol prediction model is trained based on a corresponding cohort of content items. A particular symbol prediction model of the models trained is selected based on an intrinsic characteristic of a particular content item to be losslessly compressed such as, for example, the type or file extension of the content item. The content item is then losslessly compressed based on a set of symbol predictions fed to an arithmetic coder that are generated using the particular symbol prediction model selected.

Lossless compression of a content item using a neural network trained on content item cohorts

Techniques for general-purpose lossless data compression using a neural network including compressing an original content item to a baseline lossless compressed data format. The baseline lossless compressed data format is binarized to a binarized format. The binarized format is arithmetically coded based on probability estimates from a neural network probability estimator. The neural network probability estimator generates the probability estimates for current symbols of the binarized format to be arithmetically coded based on symbols of the binarized format that have already been arithmetically coded.

Growing community of inventors

Mountain View, CA, United States of America

Daniel R Horn