SSD Installation Training Data set

SSD Installation Training data set

Transfer from http://lib.csdn.net/article/deeplearning/53859
In the home directory, get the SSD code, after the download is complete there is a Caffe folder

   git clone https://github.com/weiliu89/caffe.git
   cd caffe
   git Checkout SSD

1 2 3 Go to the downloaded Caffe directory, copy the configuration file

   Cd/home/usrname/caffe
   CP Makefile.config.example Makefile.config

1 2 1 2 compiling Caffe Trilogy

   Make-j8  //-j8 According to the processor configuration of the machine, 8 is the meaning of the eight-core processor make
   test-j8 make
   runtest-j8

1 2 3 1 2 3 additional compilation, as needed (because SSD is done with Python, compile Pycaffe)

   Make PY

preparatory work

Download the pre-training model link: http://pan.baidu.com/s/1miDE9h2 password: 0hf2, put it in the caffe/models/vggnet/directory

Download the VOC2007 and VOC2012 datasets and put them under/home/data. (Please note that the directory is changed here.)

CD..
   mkdir Data
   CD data/

1 2 3 1 2 3 download DataSet

   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

1 2 3 1 2 3 Data set decompression

   TAR-XVF Voctrainval_11-may-2012.tar
   tar-xvf voctrainval_06-nov-2007.tar
   tar-xvf VOCtest_06-Nov-2007.tar

1 2 3 1 2 3 Convert a picture to a Lmdb file for training

   CD..
   CD caffe/
   ./data/voc0712/create_list.sh  
   ./data/voc0712/create_data.sh

1 2 3 4 1 2 3 4 Here The Script implementation batch, may appear: No module named Caffe and other errors, this is because the Caffe Python environment variable is not configured, can be resolved as follows:

   echo "Export Pythonpath=/home/usrname/caffe/python" >> ~/.profile  
   source ~/.profile  
   Echo $PYTHONPATH # Check the values of environment variables

Training ModelIn the download Caffe root directory to perform the following command training, there are several. py files under EXAMPLES/SSD, the training time is longer, iteration 60,000 times, Bo Master training for one day.

   Python examples/ssd/ssd_pascal.py

1 1 Experimental Results (1) test on the picture test set

   Python examples/ssd/score_ssd_pascal.py

1 1

Using it to run over the data set, the accuracy rate can be up to 70 (2) tested on video

1 1

The use of senior pedestrian video to do the test, real-time high, but the missing rate is quite serious, it is inevitable


Of course, the direct run is his own video, want to run their own code, to first use VIM to open the file, positioning to 51 lines, modify the video path for the local video, so you can run their own video (3) on the camera head test

   Python examples/ssd/ssd_pascal_webcam.py

1 1

Bo Master transplanted to the desktop problems, has not changed the bug, changed to share to everyone later work Research SSD python source code, used to train and detect traffic signs The author given a pre-training model

If there is no good machine configuration or easier, you can use the trained model given by the author: Models trained on voc0712:ssd300 SSD500

There are other models, not listed here, interested can go to the official homepage to see, the link has been given in front

Recently has been engaged in the object detection play, before using the FASTER-RCNN, the accuracy of 73.2%, the effect is good, but the recognition speed is a bit short, I use the GPU is gtx980ti, recognition speed is about 15fps. Recently discovered SSD ( Single shot Multibox detector) This paper has a good effect and speed, I have experimented with it, the speed is indeed much faster than FASTER-RCNN. The following is a two-part introduction. The first section describes the installation of SSDs, and the second section describes how to train your own data sets based on SSDs.

1 2 3 1 2 3 the second part trains your own data sets

First of all, we might as well run a demo of the project, need to download datasets, pre-trained data sets, and so on.
Download the pre-trained model, Link: http://pan.baidu.com/s/1miDE9h2 password: 0hf2, after the download is complete to protect the existence:

caffe/models/vggnet/

1 1

Download the VOC2007 and VOC2012 datasets, placed in the/data directory:

CD data
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

tar-xvf voctrainval_11-may-2012.tar
tar-xvf VOCtrainval_06- Nov-2007.tar
TAR-XVF Voctest_06-nov-2007.tar

1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8

To create data in Lmdb format:

CD Caffe
./data/voc0712/create_list.sh
# It'll create Lmdb files for Trainval and test with encoded original IMA GE:
#   -$HOME/data/vocdevkit/voc0712/lmdb/voc0712_trainval_lmdb
#   -$HOME/data/vocdevkit/voc0712 /lmdb/voc0712_test_lmdb
./data/voc0712/create_data.sh

1 2 3 4 5 6 1 2 3 4 5 6

Training and testing:

Python examples/ssd/ssd_pascal.py