This file will tell you how to use your own dataset to train with Caffe

1. create train and val folder, which contains train and test images for example: data/train/ data/test/

2. use the label-image-for-caffe.py file to create train.txt and val.txt usage of label-image-for-caffe cmd python label-image-for-caffe.py -t 001/ 001 is the class folder for the images

3. convert_imageset use the convert_imageset.exe to convert the images usage of convert_imageset.exe cmd convert_imageset --resize_height=256 --resize_width=256 --shuffle --backend=leveldb data/train/ data/train.txt data/trainldb

4. compute_image_mean use the compute_image_mean.exe to compute the mean file usage of compute_image_mean.exe cmd compute_image_mean --backend=leveldb data/trainldb mean.binaryproto

5. some warnings the train.txt and val.txt cannot contains empty lines in it! The convert_imageset.exe may cannot work!!!