TL;DR通过共享盒子预测器并采用最大池化替代卷积之间的过程,成功降低了 Single Shot Multibox Detector 家族探测器的模型大小,同时保持相同的质量,避免了评分跨尺度的失真。
Abstract
We'd like to share a simple tweak of single shot multibox detector (SSD)
family of detectors, which is effective in reducing model size while
maintaining the same quality. We share box predictors across all scales, and
replace convolution between scales with max pooling. This has two a