Skip to first unread message
Aditya Tewari
May 27, 2015, 2:07:51 AM5/27/15
to caffe...@googlegroups.com
Trying to train a CNN with HDFData input using the recently updated caffe. The samples in HDF5 files are 3 dimensional matrix(like images). The top layer of the network is a "convolution" network, and the bottom layer is "softmaxWithLoss" which has label as the input.
On the initializtion of training I get an error saying:
"F0527 07:54:52.075439 16136 insert_splits.cpp:35] Unknown blob input data to layer 0"
The network architecture is:
name: "handPoseClassify"
layer{
name: "data"
type: "HDF5Data"
include {
phase: TRAIN
}
hdf5_data_param {
source: "data_files/training_list_h5.txt"
batch_size: 1
}
top: "data"
top: "label"
}
layer {
name: "conv1"
type: "Convolution"
convolution_param {
num_output: 20
kernel_size: 5
stride: 1
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
}
}
bottom: "data"
top: "conv1"
}
layer {
name: "pool1"
type: "Pooling"
pooling_param {
kernel_size: 2
stride: 2
pool: MAX
}
bottom: "conv1"
top: "pool1"
}
layer {
name: "conv2"
type: "Convolution"
bottom: "pool1"
top: "conv2"
convolution_param {
num_output: 50
kernel_size: 5
stride: 1
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "pool2"
type: "Pooling"
bottom: "conv2"
top: "pool2"
pooling_param {
pool: MAX
kernel_size: 2
stride: 2
}
}
layer {
name: "ip1"
type: "InnerProduct"
bottom: "pool2"
top: "ip1"
inner_product_param {
num_output: 500
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "relu1"
type: "ReLU"
bottom: "ip1"
top: "ip1"
}
layer {
name: "ip2"
type: "InnerProduct"
bottom: "ip1"
top: "ip2"
inner_product_param {
num_output: 10
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "loss"
type: "SoftmaxWithLoss"
bottom: "ip2"
bottom: "label"
top: "loss"
}
Aditya Tewari
May 27, 2015, 2:38:56 AM5/27/15
to caffe...@googlegroups.com
The shape of the sample matrix in h5 file is N*channels*row*columns (2278*2*120*165). Is it the right format or should it be row*column*channels*N?
Reply all
Reply to author
Forward
0 new messages