Extraction of spatial-temporal feature representation

A device, comprising:

• a processing unit;
• and
• a memory coupled to the processing unit and having instructions stored thereon which, when executed by the processing unit, causing the device to perform acts comprising:
• receiving an input at a first layer of a learning network;
• the input comprising a plurality of images;
• extracting, using a first unit of the first layer, first features of the plurality of images from the input in a spatial dimension, the first features characterizing a spatial presentation of the plurality of images;
• selecting a type of a connection between the first unit and a second unit of the first layer from a group comprising of a first series connection type, a second series connection type, or a parallel connection type;
• extracting second features of the plurality of images from at least one of the first features and the input in a temporal dimension, the second features at least characterizing temporal changes across the plurality of images, wherein the second unit extracts the second features, and extracting the second features comprises:
• extracting, upon determining the type of the connection is the first series connection type, at least the second features from the first features;
• extracting, upon determining the type of the connection is the second series connection type, at least the second features from the input;
• or
• extracting, upon determining the type of the connection is the parallel connection type, the second features from the input;
• and
• generating a spatial-temporal feature representation of the plurality of images based at least in part on the second features.

A computer program product being stored on a computer-readable medium and comprising machine-executable instructions which, when executed by a device, cause the device to:

• receive an input at a first layer of a learning network, the input comprising a plurality of images;
• extract, using a first unit of the first layer, first features of the plurality of images from the input in a spatial dimension, the first features characterizing a spatial presentation of the plurality of images;
• select a type of a connection between the first unit and a second unit of the first layer from a group comprising of a first series connection type, a second series connection type, or a parallel connection type;
• extract second features of the plurality of images from at least one of the first features and the input in a temporal dimension, the second features at least characterizing temporal changes across the plurality of images, wherein the second unit extracts the second features, and extracting the second features comprises:
• extracting, upon determining the type of the connection is the first series connection type, at least the second features from the first features;
• extracting, upon determining the type of the connection is the second series connection type, at least the second features from the input;
• or
• extracting, upon determining the type of the connection is the parallel connection type, the second features from the input;
• and
• generate a spatial-temporal feature representation of the plurality of images based at least in part on the second features.