Skip to main content
. 2020 Aug 31;20(17):4938. doi: 10.3390/s20174938

Figure 5.

Figure 5

Four subsets (a) ground-truth and detection results of different scenarios using (b) Faster R-CNN, (c) YOLO v3, (d) SSD. (1) Targets surrounded by crop land and residential areas in GF-1 image; (2) targets surrounded by industrial buildings in GF-1 image; (3) fragmented and serried targets in GF-2 image; (4) simple surroundings and regular targets in GF-2 image. Solid and dotted blue circles represent the existence of missing and erroneous detections, respectively.