search

shuokay / resnet

Deep Residual Learning for Image Recognition, http://arxiv.org/abs/1512.03385
85 stars 39 forks source link

input image size smaller than input shape #2

Closed pribadihcr closed 8 years ago

pribadihcr commented 8 years ago

Hi,

I have convert tiny imagenet-200 to .rec using im2rec. I got the following error: [20:04:27] src/io/iter_image_recordio.cc:212: ImageRecordIOParser: ./train.rec, use 3 threads for decoding.. [20:04:27] [20:04:27] ./dmlc-core/include/dmlc/logging.h:208: [20:04:27] src/io/./image_augmenter.h:212: Check failed: static_cast(res.rows) >= param_.data_shape[1] && static_cast(res.cols) >= param_.data_shape[2] input image size smaller than input shape [20:04:27] ./dmlc-core/include/dmlc/logging.h:208: [20:04:27] src/io/./image_augmenter.h:212: Check failed: static_cast(res.rows) >= param_.data_shape[1] && static_cast(res.cols) >= param_.data_shape[2] input image size smaller than input shape ./dmlc-core/include/dmlc/logging.h:208: [20:04:27] src/io/./image_augmenter.h:212: Check failed: static_cast(res.rows) >= param_.data_shape[1] && static_cast(res.cols) >= param_.data_shape[2] input image size smaller than input shape terminate called recursively terminate called after throwing an instance of 'dmlc::Error' terminate called recursively Aborted

many thanks for help

shuokay commented 8 years ago

imagenet-200 data size is 64*64. You should set resize large than 256 if the shape=(3,256,256). ps: the example have moved to https://github.com/shuokay/mxnet/blob/master/example/image-classification/symbol_resnet-small.py. Running this example by python train_cifar10.py --network=resnet-small --lr=0.01 --lr-factor=0.1 --lr-factor-epoch=20 --num-epochs=30 and get a final train accuracy about 91.75% and test accuracy about 84.30%