Use the faster-rcnn to do car detection

I just upload the code I change from the faster-rcnn

---

Tools to use

1. draw the RCNN network

./caffe-fast-rcnn/python/draw_net.py models/ZF/faster_rcnn_end2end/test.prototxt test.jpg

2. time the network

without python layer:

./build/tools/caffe time -model ~/fast-rcnn/models/CaffeNet/train.prototxt -gpu 0

with python layer:

see 'py-faster-rcnn-time', use 'tools/demo2.py --net zf'

---

Ignore don't care class

you have to ignore both in 'anchor_target_layer' and 'anchor_target_layer'

code in py-faster-rcnn-care

---

Train faster-rcnn in kitti dataset

code in py-faster-rcnn-blur

data set in kitti/train/label_2

---

Seperate car and occlude car to two class and train

code in py-faster-rcnn-blur3

data set in kitti/train/label_3

---

Only rpn network

code in py-faster-rcnn-conv

use the 'demo_rpn.sh' to test the network