Object localization refers to the task of identifying where an object is in an image, usually by outputting a bounding box around it, but typically in the context where the object class might be known or fixed in advance. It’s slightly different from full object detection in that often localization is used when there’s only one (or a few) object of interest in the image and you want to draw a box around it. For instance, given an image and asked “where is the cat?”, the goal is localization of the cat. In classification benchmarks like ImageNet, there was a task called “classification with localization” where a model had to not only predict the class label but also provide a bounding box for the object. Essentially, localization = detection minus the classification of multiple classes (if we assume class is known or it’s single-class detection). Techniques for localization can be simpler if it’s single-class (like class activation maps or regression of box coordinates from CNN features).
Data Selection & Data Viewer
Get data insights and find the perfect selection strategy
Learn MoreSelf-Supervised Pretraining
Leverage self-supervised learning to pretrain models
Learn MoreSmart Data Capturing on Device
Find only the most valuable data directly on devide
Learn More