sfzhang15/RefineDet

Error encountered when training my own dataset

klyuan1 opened this issue · 4 comments

Hi, Thanks for your great work!
When I was training my own data set, error happened. I only changed resize_width and resize_height in examples/refinedet/VGG16_VOC2007_512.py. Could you help? Thanks!

I0228 21:20:20.339459 8428 net.cpp:150] Setting up TL5_1_relu
I0228 21:20:20.339467 8428 net.cpp:157] Top shape: 8 256 30 45 (2764800)
I0228 21:20:20.339481 8428 net.cpp:165] Memory required for data: 27419814240
I0228 21:20:20.339484 8428 layer_factory.hpp:77] Creating layer TL5_2
I0228 21:20:20.339509 8428 net.cpp:100] Creating Layer TL5_2
I0228 21:20:20.339512 8428 net.cpp:434] TL5_2 <- TL5_1
I0228 21:20:20.339519 8428 net.cpp:408] TL5_2 -> TL5_2
I0228 21:20:20.344029 8428 net.cpp:150] Setting up TL5_2
I0228 21:20:20.344056 8428 net.cpp:157] Top shape: 8 256 30 45 (2764800)
I0228 21:20:20.344060 8428 net.cpp:165] Memory required for data: 27430873440
I0228 21:20:20.344066 8428 layer_factory.hpp:77] Creating layer P6-up
I0228 21:20:20.344075 8428 net.cpp:100] Creating Layer P6-up
I0228 21:20:20.344080 8428 net.cpp:434] P6-up <- P6_P6_relu_0_split_0
I0228 21:20:20.344086 8428 net.cpp:408] P6-up -> P6-up
I0228 21:20:20.347196 8428 net.cpp:150] Setting up P6-up
I0228 21:20:20.347208 8428 net.cpp:157] Top shape: 8 256 30 46 (2826240)
I0228 21:20:20.347213 8428 net.cpp:165] Memory required for data: 27442178400
I0228 21:20:20.347218 8428 layer_factory.hpp:77] Creating layer Elt5
I0228 21:20:20.347224 8428 net.cpp:100] Creating Layer Elt5
I0228 21:20:20.347227 8428 net.cpp:434] Elt5 <- TL5_2
I0228 21:20:20.347232 8428 net.cpp:434] Elt5 <- P6-up
I0228 21:20:20.347277 8428 net.cpp:408] Elt5 -> Elt5
F0228 21:20:20.347643 8428 eltwise_layer.cpp:34] Check failed: bottom[i]->shape() == bottom[0]->shape()
*** Check failure stack trace: ***
@ 0x7f662662edaa (unknown)
@ 0x7f662662ece4 (unknown)
@ 0x7f662662e6e6 (unknown)
@ 0x7f6626631687 (unknown)
@ 0x7f6626d85d3d caffe::EltwiseLayer<>::Reshape()
@ 0x7f6626eb8f17 caffe::Net<>::Init()
@ 0x7f6626eb9d55 caffe::Net<>::Net()
@ 0x7f6626ce7bd2 caffe::Solver<>::InitTrainNet()
@ 0x7f6626ce812c caffe::Solver<>::Init()
@ 0x7f6626ce843a caffe::Solver<>::Solver()
@ 0x7f6626ece9a3 caffe::Creator_SGDSolver<>()
@ 0x40db27 caffe::SolverRegistry<>::CreateSolver()
@ 0x408df9 train()
@ 0x40671c main
@ 0x7f6624ceff45 (unknown)
@ 0x407023 (unknown)
@ (nil) (unknown)

@klyuan1
Hi, the input size should be a multiple of 64. If you want to imput any size, you should add crop_layer before eltwise_layer to keep the input sizes of both equal.

@sfzhang15
Big thanks to you! It's really help me!

@klyuan1
Hi, the input size should be a multiple of 64. If you want to input any size, you should add crop_layer before eltwise_layer to keep the input sizes of both equal.

@sfzhang15
Could you tell me how you add the crop_layer? Really appreciated!

@china56321
Please refer to this