IIT-AFF Dataset

Anh Nguyen

Detection results using AffordanceNet on the IIT-AFF dataset.

Introduction:

The IIT-AFF dataset is a large-scale dataset that provides both the object location and its affordances. This dataset can be used in various tasks such as object detection, instance segmentation, semantic segmentation, affordance detection, etc.

The dataset has 10 object categories (bottle, bowl, cup, drill, hammer, knife, monitor, pan, racket, spatula) and 9 affordance classes (contain, cut, display, engine, grasp, hit, pound, support, w-grasp).  In particular, there are 8,835 images, 14,642 object bounding boxes (in which 7,866 bounding boxes come from the ImageNet dataset), and 24,677 affordance parts of the objects.

If you use this dataset, please cite the following paper:

Object-Based Affordances Detection with Convolutional Neural Networks and Dense Conditional Random Fields

Anh Nguyen, Dimitrios Kanoulas, Darwin G. Caldwell, Nikos Tsagarakis

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017.

Download: IIT-AFF Dataset (5.1GB) (Google Drive, One Drive)

Related Publications:

AffordanceNet: An End-to-End Deep Learning Approach for Object Affordance Detection

Thanh-Toan Do*, Anh Nguyen*, Ian Reid

ICRA 2018. (* equal contribution)

PDF Code

Detecting Object Affordances with Convolutional Neural Networks

Anh Nguyen, Dimitrios Kanoulas, Darwin G. Caldwell, Nikos Tsagarakis

IROS 2016.

PDF