matlab - Caffe using hdf5 layer and imagedata input layer together in train_val.proto -
i using hdf5 data layer reading input images(around 30000) have , metadata images.
i unable crop-flip data augmentation, since when use centre crops, storing data of 1500 images leads around 1.5 gb h5 file, total size of hdf5 dataset(30 h5 files) becomes ~40 gb, , can not use augmentation hdf5 dataset large.
so, thinking if use imagedata layer reading images , hdf5 data layer metadata, problem can solved. didn't find material on this. possible so?
yes is, thought getting error due else in proto file, below input layers proto
layer { type: "hdf5data" name: "data" top: "meta" hdf5_data_param { source: "/path/to/train.txt" batch_size: 50 } include { phase: train } } layer { name: "data" type: "imagedata" top: "x" top: "labels" include { phase: train } transform_param { mirror: true crop_size: 227 mean_file: "data/ilsvrc12/imagenet_mean.binaryproto" } image_data_param { source: "/path/to/train.txt" batch_size: 50 new_height: 256 new_width: 256 } } layer { type: "hdf5data" name: "data" top: "meta" hdf5_data_param { source: "/path/to/val.txt" batch_size: 50 } include { phase: test } } layer { name: "data" type: "imagedata" top: "x" top: "labels" include { phase: test } transform_param { mirror: false crop_size: 227 mean_file: "data/ilsvrc12/imagenet_mean.binaryproto" } image_data_param { source: "/path/to/val.txt" batch_size: 50 new_height: 256 new_width: 256 } }
Comments
Post a Comment