提出时间:2018年,a tech report
网络结构如下
backbone-Feature Extractor
darknet53
Bounding Box Prediction
Class Prediction
multilabel classification use independent logistic classifiers
Predictions Across Scales
每一个尺度都有,共三个尺度,每一个尺度3个bounding boxes.
N × N × [3 ∗ (4 + 1 + 80)] for the 4 bounding box offsets, 1 objectness prediction, and 80 class predictions.
聚类中心为:
(10 × 13), (16 × 30), (33 × 23), (30 × 61), (62 × 45), (59 × 119), (116 × 90), (156 × 198), (373 × 326).
一些不work的实验:
Anchor box x, y offset predictions.
Linear x, y predictions instead of logistic
Focal loss.
Dual IOU thresholds and truth assignment.
Faster R-CNN uses two IOU thresholds during training. If a prediction overlaps the ground truth by .7 it is as a positive example, by [.3 − .7] it is ignored, less than .3 for all ground truth objects it is a negative example. We tried a similar strategy but couldn’t get good results.
results
AP指标和SSD差不多,AP0.5和RetinaNet相当,AP0.75性能下降,APs(小尺度)好于SSD,中大尺度没有优势,之前的yolo在小尺度上是劣势