Pascal VOC is a collection of standardized image and annotation datasets for object detection, class recognition, and instance/semantic segmentation. The dataset is used to assist in standard evaluation procedures and allows for the following operations:
- The evaluation and comparison of different methods.
- A toolset for accessing datasets and annotations.
- Running challenges to gauge object class recognition.
Pascal VOC 2012 covers 20 object categories. The images in the dataset contain pixel-accurate segmentation annotations, bounding box annotations, and object class annotations, split into the following subsets:
- Training images: 1464
- Validation and private testing images: 1449
Pascal Voc has one annotation file in XML for every image in a given dataset. So far, the VOC has gone through multiple stages of development, and is being upgraded by new statistics and data every year. The most popular benchmarking combination with Pascal VOC is implementing 2007 trainval and 2012 trainval for training and 2007 for validation.
Annotation Type: object detection
Created By: Everingham, M et al
Publish Date: 2010
Dataset Size: 21,503 images