The approximate process is translated according to the GitHub tutorial HTTPS://GITHUB.COM/WEILIU89/CAFFE/TREE/SSD, and the pro-test is feasible
1, in the home directory, get the SSD code, download completed there is a Caffe folder
git clone https://github.com/weiliu89/caffe.git
cd caffe
git Checkout SSD
2. Go to the downloaded Caffe directory and copy the configuration file. If you have a configuration file installed in the Caffe directory, you can copy it.
CD /home/xxx/caffe
cp Makefile.config.example makefile.config Here write a code slice
3. Compiling Caffe Trilogy
Where 8 of-J8 represents the processor configuration of the virtual machine (I am here 2)
View command: #cat/proc/cpuinfo |grep "Cores" |uniq
Make-j2 make
test-j2 make
runtest-j2
4, download the pre-training model fully convolutional reduced (atrous) vggnet. Put it in the Caffe/models/vggnet directory
5, download VOC2007 and VOC2012 datasets, put under/home/data
CD..
mkdir Data
CD data/
6. Download Data Set
wget Http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar
wget/http Host.robots.ox.ac.uk/pascal/voc/voc2007/voctrainval_06-nov-2007.tar
wget http://host.robots.ox.ac.uk/pascal /voc/voc2007/voctest_06-nov-2007.tar
7, data set decompression, the proposed use of the command decompression, manual decompression may appear a variety of problems
TAR-XVF Voctrainval_11-may-2012.tar
tar-xvf voctrainval_06-nov-2007.tar
tar-xvf VOCtest_06-Nov-2007.tar
8. Convert image to Lmdb file for training
CD..
CD caffe/
./data/voc0712/create_list.sh
./data/voc0712/create_data.sh
If an error occurs, check to see if your Python path is configured correctly
echo "Export Pythonpath=/home/xxx/caffe_root/python" >> ~/.profile
source ~/.profile
Echo $PYTHONPATH # Check the values of environment variables
Caffe_root refers to the path of your Ssd_caffe project
9. Training model
Python examples/ssd/ssd_pascal.py