a caffe version for multiTags job
- 7 labels for all;
folder/sample001.jpg 0 1
folder/sample001.jpg 2 3 5
folder/sample001.jpg 1 6 3
folder/sample001.jpg 4 6 1 2
folder/sample001.jpg 6 3 2 1 0
...
- This script records the bug encountered during the compilation of “make”
$CAFFE_HOME/build/tools/convert_imageset -resize_height=224 -resize_width=224 / train_images.txt train_data_lmdb 7
- googlenet_train.prototxt :model prototxt
- multilabel_googlenet_iter_250000.caffemodel :model weights
- solver.prototxt :solver prototxt
- runtrain.sh :run script
- compute accuracy for the last layer
- before run the model, you need :
export PYTHONPATH=$CAFFE_HOME