An Evaluation of Deep Learning Methods for Small Object Detection

Figure 2

The visualization of detectors with the strongest backbones on subsets of PASCAL, VOC_MRA_0.58, VOC_MRA_10, VOC_MRA_20, and VOC_WH_20 in order, respectively: (a) YOLO Darknet-53; (b) Faster RCNN ResNeXT-101-64 × 4d-FPN; (c) RetinaNet ResNeXT-101-64 × 4d-FPN; (d) Fast RCNN ResNeXT-101-64 × 4d-FPN.

(a)

(b)

(c)

(d)