Generating a data structure for specifying visual data sets

A computer-implemented method for training a computer vision model, comprising:

• obtaining a further computer vision model configured to perform a computer vision function of characterising elements of observed scenes;
• obtaining a set of training data by:
• generating a data structure, including:
• obtaining a computer vision model configured to perform a computer vision function of characterizing elements of observed scenes;
• obtaining a first visual parameter set including a plurality of initial visual parameters, wherein an item of visual data provided based on an extent of the at least one initial visual parameter is capable of affecting a classification or regression performance of the computer vision model;
• providing a visual data set including a subset of items of visual data compliant with the first visual parameter set, and a corresponding subset of items of groundtruth data;
• applying the subset of items of visual data to the computer vision model to obtain a plurality of performance scores characterizing performance of the computer vision model when applied to the subset of items of visual data of the visual data set, using the corresponding groundtruth data;
• performing a sensitivity analysis of the plurality of performance scores over a domain of the first visual parameter set;
• generating a second visual parameter set including at least one updated visual parameter, wherein the second visual parameter set includes at least one initial visual parameter modified based on the outcome of the sensitivity analysis to provide the at least one updated visual parameter;
• generating the data structure including at least one language entity based on the visual parameters of the second visual parameter set, thus providing a semantic mapping to the visual parameters of the second visual parameter set;
• and
• outputting the set of visual data and corresponding items of groundtruth data as the training data set;
• and
• training the computer vision model using the set of training data.

An apparatus for generating a data structure comprising a plurality of language entities defining a semantic mapping of visual parameters to a visual parameter space, comprising:

• an input interface;
• a processor;
• a memory;
• and
• an output interface;
• wherein the input interface is configured to obtain a computer vision model configured to perform a computer vision function of characterizing elements of observed scenes, and to obtain a first visual parameter set including a plurality of initial visual parameters, wherein an item of visual data provided based on an extent of the at least one initial visual parameter is capable of affecting a classification or regression performance of the computer vision model, and
• wherein the processor is configured to:
• provide a visual data set including a subset of items of visual data compliant with the first visual parameter set, and a corresponding subset of items of groundtruth data,
• apply the subset of items of visual data to the computer vision model to obtain a plurality of performance scores characterizing the performance of the computer vision model when applied to the subset of items of visual data of the visual data set, using the corresponding groundtruth data,
• perform a sensitivity analysis of the plurality of performance scores over a domain of the first visual parameter set,
• generate a second visual parameter set including at least one updated visual parameter, wherein the second visual parameter set includes at least one initial visual parameter modified based on the outcome of the sensitivity analysis to provide the at least one updated visual parameter, and
• generate a data structure comprising at least one language entity based on the visual parameters of the second visual parameter set, thus providing a semantic mapping to visual parameters of the second visual parameter set.

A non-transitory computer readable medium on which is stored a computer program for generating a data structure including a plurality of language entities defining a semantic mapping of visual parameters to a visual parameter space, the computer program, when executed by a processor, causing the processor to perform the following steps:

• obtaining a computer vision model configured to perform a computer vision function of characterizing elements of observed scenes;
• obtaining a first visual parameter set including a plurality of initial visual parameters, wherein an item of visual data provided based on an extent of the at least one initial visual parameter is capable of affecting a classification or regression performance of the computer vision model;
• providing a visual data set including a subset of items of visual data compliant with the first visual parameter set, and a corresponding subset of items of groundtruth data;
• applying the subset of items of visual data to the computer vision model to obtain a plurality of performance scores characterizing performance of the computer vision model when applied to the subset of items of visual data of the visual data set, using the corresponding groundtruth data;
• performing a sensitivity analysis of the plurality of performance scores over a domain of the first visual parameter set;
• generating a second visual parameter set including at least one updated visual parameter, wherein the second visual parameter set includes at least one initial visual parameter modified based on the outcome of the sensitivity analysis to provide the at least one updated visual parameter;
• and
• generating a data structure including at least one language entity based on the visual parameters of the second visual parameter set, thus providing a semantic mapping to the visual parameters of the second visual parameter set.