How to change image size to 1*128 (height * width) for CaffeNet?

[Chias JaJa]

#!/usr/bin/env sh

MY=examples/mytry

echo "Create train lmdb.."

rm -rf $MY/img_train_lmdb

build/tools/convert_imageset \

--shuffle \

--resize_height=50 \

--resize_width=50 \

/home/chias/桌面/caffe-master/data/image/ \

$MY/train.txt \

$MY/img_train_lmdb

echo "Create test lmdb.."

rm -rf $MY/img_test_lmdb

build/tools/convert_imageset \

--shuffle \

--resize_width=50 \

--resize_height=50 \

/home/chias/桌面/caffe-master/data/image/ \

$MY/test.txt \

$MY/img_test_lmdb

echo "All Done.."

if I set resize to 1*128 (height * width), I can get error .....

I0331 03:49:30.395200 17970 layer_factory.hpp:76] Creating layer data

I0331 03:49:30.396656 17970 net.cpp:106] Creating Layer data

I0331 03:49:30.396710 17970 net.cpp:411] data -> data

I0331 03:49:30.396767 17970 net.cpp:411] data -> label

I0331 03:49:30.396782 17972 db_lmdb.cpp:38] Opened lmdb examples/mytry/img_train_lmdb

I0331 03:49:30.396808 17970 data_transformer.cpp:25] Loading mean file from: examples/mytry/mean.binaryproto

I0331 03:49:30.396989 17970 data_layer.cpp:41] output data size: 256,3,1,128

I0331 03:49:30.400944 17970 net.cpp:150] Setting up data

I0331 03:49:30.400993 17970 net.cpp:157] Top shape: 256 3 1 128 (98304)

I0331 03:49:30.401006 17970 net.cpp:157] Top shape: 256 (256)

I0331 03:49:30.401012 17970 net.cpp:165] Memory required for data: 394240

I0331 03:49:30.401080 17970 layer_factory.hpp:76] Creating layer conv1

I0331 03:49:30.401098 17970 net.cpp:106] Creating Layer conv1

I0331 03:49:30.401108 17970 net.cpp:454] conv1 <- data

I0331 03:49:30.401124 17970 net.cpp:411] conv1 -> conv1

F0331 03:49:30.402144 17970 blob.cpp:32] Check failed: shape[i] >= 0 (-1 vs. 0) 

*** Check failure stack trace: ***

    @     0x7fca2bf9ddaa  (unknown)

    @     0x7fca2bf9dce4  (unknown)

    @     0x7fca2bf9d6e6  (unknown)

    @     0x7fca2bfa0687  (unknown)

    @     0x7fca2c4bfc4b  caffe::Blob<>::Reshape()

    @     0x7fca2c5abc6a  caffe::BaseConvolutionLayer<>::Reshape()

    @     0x7fca2c51a4f6  caffe::Net<>::Init()

    @     0x7fca2c51b5a5  caffe::Net<>::Net()

    @     0x7fca2c4e758a  caffe::Solver<>::InitTrainNet()

    @     0x7fca2c4e879c  caffe::Solver<>::Init()

    @     0x7fca2c4e8aa9  caffe::Solver<>::Solver()

    @     0x7fca2c4f6023  caffe::Creator_SGDSolver<>()

    @           0x410f66  caffe::SolverRegistry<>::CreateSolver()

    @           0x40a2fb  train()

    @           0x4081e1  main

    @     0x7fca2acfeec5  (unknown)

    @           0x40899d  (unknown)

    @              (nil)  (unknown)

How to change image size to 1*128 (height * width) for CaffeNet?

[Ahmed Ibrahim]

Conv. layers gets smaller and smaller. you have to create an architecture that can handle such dimension , i have no idea how this can be done , why would you want to do that anyway ?