摘要: tf-faster-rcnn github:https://github.com/endernewton/tf-faster-rcnn backbone,例如vgg,conv层不改变feature大小,pooling层输出(w/2, h/2),有4个pooling所以featuremap变为原图1/ 阅读全文
posted @ 2018-02-28 17:08 demianzhang 阅读(3799) 评论(0) 推荐(0) 编辑