Multi-label Classification Using RetinaNet Model

Abstract

Object detection and classification have several versions which include image classification, object localization within an image, to multiple object detection and localization. Object recognition or classification has moved further to multi-label classification of an image and of an object. In present state of art models, multi-label image classification or a single label for object localization in images is available. Multiple labels for object classification and localization are not available. We present a framework accomplished by the study and customization of the RetinaNet architecture, to do object classification and localization with multiple labels for multiple classes of objects.