panda-lab / faceboxes Goto Github PK

这个优化版和论文不一样啊，没有prior box加密的部分。倒是你提供的prototxt里有多min_size，prior_box_layer.cpp里也有对应的改变，感觉那个优化版好奇怪啊。。。

在fddb和wider face上结果是多少？另外solver怎么配的？谢谢！

hi，我最近也想复现这个论文，请问一下你复现的结果90%是在fp=多少的时候呢，有具体的roc曲线吗，十分感谢！

Hi! Thanks for the provided code first of all!
I guess there is a mistake in "readme.md" file, you mentioned about 90% wider face dataset accuracy, but I tested caffemodel in this repository, model in https://github.com/lippman1125/faceboxes_lqy/tree/ssd/examples/faceboxes/demo and finally trained my own model. Max accuracy reached on WiderFace was about 62% (VOC metric), at the same time it has shown about 90% (VOC) for FDDB with 1024 input.

你好，请问下你这个复现结果和论文相比，出入大么

(aspect_ratios_.size()+1) is a typo ? I think right equation is : num_priors_ += aspect_ratios_.size() * (pow(densitys_[i],2)-1)

您好：
我在ssd的caffe分支中，替换掉原ssd的prior_box_layer.cpp文件，编译没有问题，但是在执行"caffe time"测试耗时时，出异常了：
I1016 15:07:06.467943 7093 net.cpp:283] Network initialization done.
I1016 15:07:06.468346 7093 caffe.cpp:355] Performing Forward
*** Aborted at 1508137626 (unix time) try "date -d @1508137626" if you are using GNU date ***
PC: @ 0x7f091b5cd78d caffe::PriorBoxLayer<>::Forward_cpu()
*** SIGSEGV (@0x0) received by PID 7093 (TID 0x7f091bb86a40) from PID 0; stack trace: ***
@ 0x7f0919952cb0 (unknown)
@ 0x7f091b5cd78d caffe::PriorBoxLayer<>::Forward_cpu()
@ 0x7f091b6f02da caffe::Net<>::ForwardFromTo()
@ 0x7f091b6f04b7 caffe::Net<>::Forward()
@ 0x4090ea time()
@ 0x4062ac main
@ 0x7f091993df45 (unknown)
@ 0x406bb3 (unknown)
@ 0x0 (unknown)
*** Error in `./build/tools/caffe': malloc(): memory corruption: 0x0000000002052d00 ***
Aborted (core dumped)
应该是prior_box_layer.cpp引入的；
是我使用的caffe分支有问题，还是别的问题呢？望解答，谢谢！

大神 请问我在训练时输出多个detection_eval值，个数也不固定，原因是什么那？
例如：I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2853: detection_eval = 0
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2854: detection_eval = 0.98
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2855: detection_eval = 6.86
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2856: detection_eval = 0.98
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2857: detection_eval = 1.34987
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2858: detection_eval = 0
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2859: detection_eval = 0.98
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2860: detection_eval = 6.72
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2861: detection_eval = 0.96
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2862: detection_eval = 1.31655
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2863: detection_eval = 0
I1112 19:23:18.045190 8388 solver.cpp:425] Test net output #2864: detection_eval = 0.96
I1112 19:23:19.592190 8388 solver.cpp:243] Iteration 0, loss = 7.11032
I1112 19:23:19.592190 8388 solver.cpp:259] Train net output #0: mbox_loss = 7.

我发现Pool2输出的尺寸是31x31，请问这没问题吗，会不会是32x32？