Media content enrichment using an adapted object detector

A system comprising:

• at least one computing device comprising one or more processors to execute and memory to store instructions to:
• obtaining logic executed by the processor for obtaining a query comprising at least one term to retrieve a plurality of first media content items;
• retrieving logic executed by the processor for retrieving the plurality of first media content items using the query and an existing association between the at least one term and each first media content item of the plurality indicating that each of the first media content items of the plurality depicts an object belonging to a first visual object class;
• generating logic executed by the processor for generating an adapted visual object detector using an existing visual object detector used to detect a second visual object class different from the first visual object class, the adapted visual object detector being generated by retraining the existing visual object detector using the first media content items depicting an object belonging to the first visual object class to detect objects belonging to the first visual object class, the adapted visual object detector detects a single visual object class that is different from the visual object class detected by the existing visual object detector, the existing visual object detector comprises the cascade of classifiers to detect a plurality of visual features associated with one or more visual objects belonging to the second visual object class, the generating logic further comprising replacing logic for replacing a subset of the classifiers in a cascade of classifiers with a probabilistic generative model trained using a non-negative factorization of a matrix constructed using the first media content items;
• and
• analyzing logic executed by the processor for analyzing a second media content item of a plurality of second media content items using the adapted visual object detector to determine whether the second media content item depicts a visual object belonging to the first visual object class.

A computer readable non-transitory storage medium for tangibly storing thereon computer readable instructions that when executed cause at least one processor to:

• obtain a query comprising at least one term to retrieve a plurality of first media content items;
• retrieve the plurality of first media content items using the query and an existing association between the at least one term and each first media content item of the plurality indicating that each of the first media content items of the plurality depicts an object belonging to a first visual object class;
• generate an adapted visual object detector using an existing visual object detector used to detect a second visual object class different from the first visual object class, the adapted visual object detector being generated by retraining the existing visual object detector using the first media content items depicting an object belonging to the first visual object class to detect objects belonging to the first visual object class, the adapted visual object detector detects a single visual object class that is different from the visual object class detected by the existing visual object detector, the existing visual object detector comprises a cascade of classifiers to detect a plurality of visual features associated with one or more visual objects belonging to the second visual object class, the instructions to generate further comprising instructions to replace a subset of the classifiers in the cascade of classifiers with a probabilistic generative model trained using a non-negative factorization of a matrix constructed using the first media content items;
• and
• analyze a second media content item of a plurality of second media content items using the adapted visual object detector to determine whether the second media content item depicts a visual object belonging to the first visual object class.