TensorFlow-R-FCN-for-primary-students

1.Introduction

R-FCN Tensorflow version for primary students

Mainly refer to https://github.com/auroua/tf_rfcn and https://github.com/endernewton/tf-faster-rcnn

Thanks for auroua and endernewton's help.

This code is for primary students to use.

What is different from theirs?

1. Support Python3 Train and Test

2. Support Chinese Language of detecting result

3. More easy to use and small changes

Requirements

System: Unix or Unix-Like Platrform
(Have only tested on linux-ubuntu, other unix system should be supported in theory)
Python3
Tensorflow 1.8+
OpenCV-Python
CUDA
CUDNN

Usage

You need to download weights first. See the following Models Download. Inference

    git clone https://github.com/engineer1109/TensorFlow-R-FCN-for-primary-students.git  
    cd lib && make
    cd ..
    sh run.sh 

Train

    sh train.sh

Tensorboard viewer

    sh tensorboard.sh

Dataset of Pascal Voc

Pascal Voc 2007+2012
Unzip it to ./data/VOCdevkit2007/VOC2007

Models Download

Main Model Unzip it to ./output

Download	Path
Baidu Yun	https://pan.baidu.com/s/1-zAgNRavUJGX5ggmSH-TEQ Code: b4vt
Google	https://drive.google.com/open?id=1iWMqZAX3wh-XN0o5pIGClvIx8wMPWWFT

Pretrained Model Unzip it to ./data/imagenet_weights

Download	Path
Baidu Yun	https://pan.baidu.com/s/1oSZMlwOaIFOuqrMeSIY-PQ Code: qvos
wget	http://download.tensorflow.org/models/resnet_v1_101_2016_08_28.tar.gz

For wget, you need to change filename to res101.ckpt

    wget -v http://download.tensorflow.org/models/resnet_v1_101_2016_08_28.tar.gz
    tar -xzvf resnet_v1_101_2016_08_28.tar.gz
    mv resnet_v1_101.ckpt res101.ckpt

This is the first complete version. V1.0