1. Database for 3D object recognition in cluttered scenes : There are 50 different scenes scanned with the Minolta scanner. Each scene contains 4 to 5 objects placed randomly causing clutter and occlusions. This database is also available from Point Cloud Library PCL and 3D Keypoint Detection Benchmark.

2. Database for 3D surface registration (3D modeling) : Multiple views (2.5D scans) of objects. Useful for testing automatic correspondence/registration and surface integration algorithms. The full 3D models are available in the object recognition database above.

3. FACE DATABASE (3D scans and 2D images under different illuminations): Frontal face images under varying illumination (from an LCD) and the corresponding 3D face model (acquired with Minolta laser scanner).  The facial images are cropped and normalized. This database is AVAILABLE ONLY FOR THE PURPOSE OF FACE RECOGNITION AND RECONSTRUCTION RESEARCH. Commercial or any other use of the database is not allowed. By downloading the database, you agree that you will only use the database for the allowed purpose.

Subjects 1 to 10  (52 MB)
Subject 11 to 30 (73 MB
Subject 31 to 40 (44 MB)
Subject 41 to 50 (46 MB)
Subject 51 to 60 (50 MB)
Subject 61 to 70 (64 MB)
Subject 71 to 80 (53 MB)
Subject 81 to 90 (49 MB)
Subject 91 to 106 (61 MB)
Average 3D Face (Mean 3D Face) 
rendering [Download PLY]

Note: Images/3D models of some subjects are missing because they did not allow their images to be distributed. Please cite the following papers if you use the database:

Ajmal Mian, "Illumination Invariant Recognition and 3D Reconstruction of Faces using Desktop Optics", Optics Express, vol. 19(8), pp. 7491--7506, 2011. [pdf]

Ajmal Mian, "Shade Face: Multiple Image based 3D Face Recognition", 3D Digital Imaging and Modeling held jointly with ICCV, 2009. [pdf] Copyright © IEEE.

4. AIRCRAFT TRACKING DATABASE : This database can be used for research purposes only. Commercial use of the database is not allowed. There are 14 video sequences of small and large aircrafts. Variations include clouds, background buildings, occlusions by trees, specularities, camera movements, multiple aircrafts, smoke, birds and aerobatics. Please cite the following paper if you use this database.

AirCraft_tracking.zip (38 MB)

Ajmal Mian, "Realtime Visual Tracking of Aircrafts'', Digital Image Computing: Techniques and Applications (DICTA), pp. 351-356, 2008. [pdf] Copyright © IEEE.

5. HYPERSPECTRAL VIDEO : This database has a 31 frame hyperspectral video. Each frame has 33 bands from 400nm to 720nm wavelength with a 10nm step. Database is available for research only. Please cite the following paper if you use the database.

Hyperspectral Video (171 MB)

Ajmal Mian and Richard Hartley, "Hyperspectral Video Restroration using Optical Flow and Sparse Coding", Optics Express, vol. 20(10), pp. 10658-10673, 2012. [pdf]

6. HYPERSPECTRAL DOCUMENT ANALYSIS : This database contains 33 band hyperspectral images of documents handwritten by 7 subjects using 5 different blue and 5 different black inks. It can be used for ink mismatch detection or ink segmentation. Please cite the following paper if you use the database.

Hyperspectral Document Images (136 MB)

Z. Khan, F. Shafait and Ajmal Mian, "Hyperspectral Imaging for Ink Mismatch Detection," Int. Conf. on Document Analysis and Recognition (ICDAR), 2013.

7. HYPERSPECTRAL REFLECTANCE/ILLUMINANT RECOVERY : UWA Multi-illuminant Hyperspectral Scene Database v1.0. This database contains 33 band hyperspectral images (400-720nm with 10nm step) of blocks of different colours and shapes under different illuminatins. Please cite the following paper if you use the database.

UWA-MIHS-v1 (70 MB)

Z. Khan, F. Shafait and A. Mian, "Adaptive Spectral Reflectance Recovery Using Spatio-Spectral Support from Hyperspectral Images", Int. Conference on Image Processing (ICIP), 2014

8. UWA3D Activity Dataset (ECCV 2014): This dataset contains Kinect data (3D) for human action recognition. There are 30 actions performed by 10 actors. The data is divided into multiple files below for downloading.

UWA3D Activity-1
UWA3D Activity-2
UWA3D Activity-3
UWA3D Activity-4
UWA3D Activity-5
UWA3D Activity-6

The above data is only the front view part. Cross view data will be posted soon. Cite the following paper if you use the database.

H. Rahmani, A. Mahmood, D. Huynh and Ajmal Mian, "HOPC: Histogram of Oriented Principal Components of 3D Point Clouds for Action Recognition", European Conference on Computer Vision (ECCV), 2014. [pdf] [MATLAB code]