Extrinsic Least Squares Regression with Closed-Form Solution on Product Grassmann Manifold for Video-Based Recognition

Figure 3

Examples from the USCD dataset: “light”; “medium”; “heavy.”