Systems and methods for similarity-based universal object-detection for joint identification

A system comprising:

• a processor;
• and
• a memory having stored thereon instructions that are executable by the processor to cause the system to perform operations comprising:
• periodically training, by the system, a machine learning model for classifying images as containing at least one of a predefined object or a generic object on a corpus of objects, wherein:
• the corpus of objects is partially labeled with specific categories;
• and
• training the machine learning model comprises training the machine learning model to detect objects that are not labeled as belonging to a generic category that comprises the specific categories;
• receiving live updates, by the system, to a library of images of objects that have not been incorporated into the periodic training of the machine learning model;
• receiving, by the system, a plurality of images subject to classification;
• classifying, by the system, a first portion of the plurality of images as containing a predefined object;
• classifying, by the system, a second portion of the plurality of images as containing a generic object;
• and
• matching, for the second portion of the plurality of images, each image within the plurality of images to one or more images within the library of images.

A computer-implemented method comprising:

• receiving, by a computing system, an image to categorize;
• providing, by the computing system, the image as input to a pretrained machine learning model;
• determining, by the computing system, based on output from the pretrained machine learning model, that the image corresponds to a generic category;
• performing, by the computing system and based on determining that the image corresponds to a generic category, a similarity search between the image and a library of images;
• and
• categorizing, by the computing system, the image based at least in part on the similarity search;
• and
• training, by the computing system, the pretrained machine learning model on a corpus of objects, wherein:
• the corpus of objects is partially labeled with specific categories;
• and
• training the machine learning model comprises training the machine learning model to detect objects that are not labeled as belonging to a generic category that comprises the specific categories.