search

daijifeng001 / R-FCN

R-FCN: Object Detection via Region-based Fully Convolutional Networks
Other
1.24k stars 542 forks source link

crash when train R-FCN with Resnet50/101 #48

Open ouxinyu opened 7 years ago

ouxinyu commented 7 years ago
  1. The script_rfcn_demo.m is work well with Resnet101
  2. But the matlab go crash when running with Resnet50/101, the code is down at line 117 "caffe_solver.step(1)" of rfcn_train.m, without any message. It seems to be the reason for bn layers.
  3. I wrote a solver with VGG16, everything is ok

Can you help me ?Thank you! @daijifeng001

xiaoxiongli commented 7 years ago

please enable the caffe log in the matlab to see what happend, it will have a log file to give you some hint about the error.

ouxinyu commented 7 years ago

@xiaoxiongli 请问你有没有更改resnet版本的faster RCNN（主要是RPN部分），是否能共享一下代码？因为R-FCN没有提供RPN 谢谢！

xiaoxiongli commented 7 years ago

@ouxinyu I recommend you that use the R-FCN's python version, it is trained in a end2end method(not 4-stage training), the end2end training method is 1.5x faster than 4-stage training. and it is more simpler: only write the network down as a single model and just train it. So i am focus on it now and i am learning Python now. I got a mAP=97.2(faster-rcnn is about mAP=90) in my own training data and test data(229 category logo, about 3w images) with our 4-stage(alternating optimization) training. R-FCN is really good. For the code commit to github or release to anyone else, I need some discuss with my CEO, So Please try the RFCN's Python version first. bless.

limin1130 commented 7 years ago

@xiaoxiongli Does the python version only work in Linux ? what about windows ?

limin1130 commented 7 years ago

@ouxinyu have you solve your problem ? I met it also and the PC shutdown suddenly.