Method and system for providing visual explanations for image analytics decisions

A system for providing visual explanations for image analytics decisions, the system comprising:

• a processor;
• and
• a memory communicatively coupled to the processor, wherein the memory stores processor-executable instructions, which, on execution, cause the processor to:
• extract a set of local features from each of a plurality of image instances using a deep learning (DL) model, wherein the plurality of image instances is associated with an image of a given class to be analysed, and wherein the image is classified into the given class based on a set of global features;
• determine a feature list by aggregating the set of local features from each of the plurality of image instances;
• generate a two-dimensional (2D) pixel map based on the feature list, and wherein generating the 2D pixel map comprises determining pixel characteristics for each of the set of local features from each of the plurality of image instances;
• superimpose the 2D pixel map of aggregated features on each of the plurality of image instances;
• and
• provide a visual explanation for an image analytics decision on one or more of the plurality of image instances based on superimposition, wherein providing the visual explanation comprises visually indicating a plurality of mark-ups on the one or more of the plurality of image instances based on a degree of match between each of the one or more of the plurality of image instances and the 2D pixel map.

A non-transitory computer-readable medium storing computer-executable instructions for providing visual explanations for image analytics decisions, the computer-executable instructions configured for:

• extracting a set of local features from each of a plurality of image instances using a deep learning (DL) model, wherein the plurality of image instances is associated with an image of a given class to be analysed, and wherein the image is classified into the given class based on a set of global features;
• determining a feature list by aggregating the set of local features from each of the plurality of image instances;
• generating a two-dimensional (2D) pixel map based on the feature list, and wherein generating the 2D pixel map comprises determining pixel characteristics for each of the set of local features from each of the plurality of image instances;
• superimposing the 2D pixel map of aggregated features on each of the plurality of image instances;
• and
• providing a visual explanation for an image analytics decision on one or more of the plurality of image instances based on superimposition, wherein providing the visual explanation comprises visually indicating a plurality of mark-ups on the one or more of the plurality of image instances based on a degree of match between each of the one or more of the plurality of image instances and the 2D pixel map.