udn_extend

pedestrian detection on Caltech dataset

1. This model can achieve 9.7% miss rate on refined Caltech reasonable test sets.

Perparation in advance:

1. download the test images(Caltech test sets, 4024 frames).

The testing step:

1. download model and prototxt from baidu cloud disk.

2. unzip the zip file in the root directory.

3. run experiments/script_rpn_rcnn_new.m

If you have some problems to run this code, please contact ZHOU Hui(smarthuizhou@gmail.com).