Exploring the datasets

The datasets available for object localization and detection are many. In this section, we will explore the datasets that are used by the research community to evaluate the algorithms. There are datasets with a varying number of objects, ranging from 20 to 200 annotated in these datasets, which makes object detection hard. Some datasets have too many objects in one image compared to other datasets with just one object per image. Next, we will see the datasets in detail.

ImageNet dataset

ImageNet has data for evaluating classification, localization, and detection tasks. Similar to classification data, there are 1,000 classes for localization tasks. The accuracy is calculated based on the top five detections. There will be at least one bounding box in all the images. There are 200 objects for detection problems with 470,000 images, with an average of 1.1 objects per image. 

PASCAL VOC challenge

The PASCAL VOC challenge ran from 2005 to 2012. This challenge was considered...