PASCAL VOC Dataset

Submitted by on Aug 25 2014 } Suggest Revision
By:
Project: http://pascallin.ecs.soton.ac.uk/challen...
From: University of Southampton
Paper:
Resource Type:
Data
License:
Other/Unknown
Language:
N/A
Data Format:
Images

Description

The PASCAL VOC project provides standardised image data sets for object class recognition and a common set of tools for accessing the data sets and annotations, and enables evaluation and comparison of different methods.
Categorized in: Sensing and Vision | Image Understanding
Post comment
Cancel
External User
2014-11-06 17:52:29

Standard resource for object recognition, action recognition, segmentation, etc. Very well maintained and reliably annotated. Frank Keller

External User
2014-11-05 18:17:32

Well respected and studied dataset. Well organized and relevant challenges. Brendan Morris

External User
2014-10-31 08:32:59

It is a well-established and widely-used benchmark. Cong Yao

External User
2014-10-30 08:28:04

Currently, it is the best computer vision dataset for evaluating object detection algorithms. It has had a long history and has been instrumental in greatly improving the state-of-the-art of object detection. Alexei Efros

External User
2014-10-30 05:07:28

Currrently, in my experience, most of machine learning algorithms do not performe well on real world problems. VOC with supplied SIFT feature would supply more challange testbase for many machine learning problems including feature learning, optimization, classification. On the other side, if more meta-information of VOC is supplied, VOC may be more attractive. Junbiao Pang

External User
2014-10-30 04:00:15

A nice dataset that has really helped to advance the visual recognition field by standardizing the benchmarking of different algorithms. Sibt ul Hussain

External User
2014-10-29 17:06:45

A standard, well thought out benchmark and a well-collected set of annotations and images Bharath Hariharan

External User
2014-10-29 16:43:21

It is a very nice dataset to benchmark algorithms in object detection. Recently the ImageNet detection challenge has added large-scale flavor to this problem, but VOC still remains a useful dataset to test things quickly and compare against baseline approaches. Yangqing Jia

External User
2014-10-29 16:24:29

Well annotated, good dataset for researchers starting work in classification and detection. State-of-art on this dataset is widely known and understood. Kevin McGuinness

External User
2014-10-29 16:07:52

The PASCAL VOC Dataset provides a set of challenges for the tasks of whole image classification, Object localisation and classification (detection), and object-class segmentation (image parsing, semantic segmentation). The ground truth labels are perhaps the most precise out of all publicly available computer vision data-sets, especially for bounding boxes and segmentation data. There is an overlap of the data such that all segmented images also have bounding box annotation and all images annotated with bounding boxes have classification labels, such that you can try joint approaches. The data is carefully created by experts to be challenging for current computer vision algorithms, such that it is not hopelessly impossible, but rather on the tantalizing difficult side. I great data-set to work with over all, and a must for object detection evaluation. Paul Sturgess

External User
2014-10-29 15:59:36

I helped create it :) John Winn

External User
2014-10-29 15:56:21

It has enabled unprecedented progress in object recognition. Pablo Arbelaez