Cambridge Hand Gesture Data set  

  • Cambridge-Gesture data base is now publicly available at

Set1, Set2, Set3, Set4, Set5

The size of the data set is about 1GB. The citation paper for this data set is

T-K. Kim and R. Cipolla, Canonical Correlation Analysis of Video Volume Tensors for Action Categorization and Detection, IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 31(8):1415-1428, 2009. Or

T-K. Kim, S-F. Wong and R. Cipolla, Tensor Canonical Correlation Analysis for Action Classification, In Proc. of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Minneapolis, MN, 2007.

  • The data set consists of 900 image sequences of 9 gesture classes, which are defined by 3 primitive hand shapes and 3 primitive motions (see Figure 1). Therefore, the target task for this data set is to classify different shapes as well as different motions at a time.

Figure 1.
Hand-Gesture Database.
9 different gesture classes are generated by 3 different primitive shapes and motions.

  • Each class contains 100 image sequences (5 different illuminations x 10 arbitrary motions x 2 subjects). Each sequence was recorded in front of a fixed camera having roughly isolated gestures in space and time. Thus, fairly large intra-class variations in spatial and temporal alignment is reflected to the data set. See Figure 2 for typical sample sequences of the 9 classes and Figure 3 for 5 different illumination prototypes.

Figure 2. Sample sequences of the 9 gesture classes.


Figure 3. 5 different illumination conditions in the database.