In simple words: Detection identifies objects and their locations.
Beyond classification
Image classification gives one or more categories for an entire image. Object detection predicts the category and location of each object. A bounding box is the rectangle that marks an object’s position.
Region-based detection
Region-based Convolutional Neural Network (R-CNN) first considers possible image regions and then classifies what each region contains. Later Fast R-CNN and Faster R-CNN improved the process.
See the flow
image→candidate regions→CNN features→boxes + labels
Simple example
A traffic photo contains a car, bicycle, and pedestrian. Detection gives a labelled box for each one.
Remember: Detection identifies objects and their locations.