Method for analyzing an output of an object detector

A method for analyzing an output of an object detector for a selected object of interest, the method comprising:

• selecting the object of interest in a first image;
• adding the bounded object to a plurality of proposals;
• running a first inference operation on the first image;
• running a non-max suppression (NMS) algorithm on the plurality of proposals, wherein the plurality of proposals includes the bounded object;
• preventing the NMS algorithm from discarding a proposal of the plurality of proposals that includes the object of interest;
• running a classifier and bounding box regressor on each proposal of the plurality of proposals and outputting the results;
• and
• analyzing the outputted results of the object detector.

A method for analyzing an output of an object detector for a selected object of interest, the method comprising:

• selecting the object of interest in a first image;
• drawing, by the user, a plurality of bounding boxes around the object of interest, each bounding box of the plurality of bounding boxes having a different sized margin;
• isolating each bounding box with the different sized margins from the rest of the image;
• running an inference operation for each isolated bounding box of the plurality of bounding boxes, wherein the object detector provides a plurality of proposals for each inference operation;
• running a non-max suppression (NMS) algorithm on the plurality of proposals;
• running a classifier and bounding box regressor on each proposal of the plurality of proposals and outputting the results;
• and
• analyzing the outputted results of the object detector.