Fig. 6.

Example images from each scene category and their corresponding multiple object labels in UCM multi-label dataset. Each image is $256\times 256$ pixels with a spatial resolution of one foot, and their scene and object labels are introduced: (a) Agricultural: field and tree. (b) Airplane: airplane, bare soil, car, grass and pavement. (c) Baseball diamond: bare soil, building, grass, and pavement. (d) Beach: sand and sea. (e) building: building, car, and pavement. (f) Chaparral: bare soil and chaparral. (g) Dense residential: building, car, grass, pavement, and tree. (h) Forest: building, grass, and tree. (i) Free way: bare soil, car, grass, pavement, and tree. (j) Golf course: grass, pavement, sand, and tree. (k) Harbor: dock, ship, and water. (l) Intersection: building, car, grass, pavement, and tree. (m) Medium residential: building, car, grass, pavement, and tree. (n) Mobile home park: bare soil, car, grass, mobile home, pavement, and tree. (o) Overpass: bare soil, car, and pavement. (p) Parking lot: car, grass, and pavement. (q) River: grass, tree, and water. (r) Runway: grass and pavement. (s) Sparse residential: bare soil, building, car, grass, pavement, and tree. (t) Storage tank: bare soil, pavement, and tank. (u) Tennis court: bare soil, court, grass, and tree.