LR-FTI-FDet

This repository includes the code for the paper "A Unified Light Framework for Real-time Fault Detection of Freight Train Images". The code for LR-FTI-FDet is based on Soft-NMS and Faster RCNN.

Requirements: software
Requirements: hardware
Basic installation
Demo
Beyond the demo: training and testing
Usage

Requirements: software

NOTE If you are having issues compiling and you are using a recent version of CUDA/cuDNN, please consult this issue for a workaround

Requirements for Caffe and pycaffe (see: Caffe installation instructions)

Note: Caffe must be built with support for Python layers!

# In your Makefile.config, make sure to have this line uncommented
WITH_PYTHON_LAYER := 1
# Unrelatedly, it's also recommended that you use CUDNN
USE_CUDNN := 1

Python packages you might not have: cython, python-opencv, easydict

Requirements: hardware

For training the end-to-end version of LR-FTI-FDet with the backbone RFDNet, 2G of GPU memory is sufficient (using CUDNN)

Installation (sufficient for the demo)

Clone the LR-FTI-FDet repository

# Make sure to clone with --recursive
git clone --recursive https://github.com/tmpyzhang/LR-FTI-FDet.git

We'll call the directory that you cloned LR-FTI-FDet into ROOT`
Build the Cython modules
```
cd $ROOT/lib
make
```

Build Caffe and pycaffe

cd $ROOT/caffe
# Now follow the Caffe installation instructions here:
#   http://caffe.berkeleyvision.org/installation.html

# If you're experienced with Caffe and have all of the requirements installed
# and your Makefile.config in place, then simply do:
make -j8 && make pycaffe

Download pre-computed LR-FTI-FDet detectors (ach already in the folder output\) These models were trained on six fault datasets in our paper.

Demo

To run the demo

cd $ROOT
./tools/demo_tfds.py

The demo performs detection using a RFDNet network trained for detection on Angle cock dataset.

Beyond the demo: installation for training and testing models

Build your own dataset containing the training, validation, test data. It should have this basic structure

$data/                                  # data
$data/VOCdevkit2007/                    # following the format of VOC
$data/VOCdevkit2007/VOC2007/            # image sets, annotations, etc.

Download pre-trained ImageNet models

Pre-trained ImageNet models can be downloaded for the networks described in the paper: RFDNet. For convenience, we renamed it to "VGG16.v2". (already in the folder data\imagenet_models)

Usage

To train and test a LR-FTI-FDet detector using the approximate joint training method, use experiments/scripts/faster_rcnn_end2end.sh. Output is written underneath $ROOT/output.

cd $ROOT
./experiments/scripts/faster_rcnn_end2end.sh [GPU_ID] [NET] [--set ...]
# GPU_ID is the GPU you want to train on
# NET in {VGG16} is the network arch to use
# --set ... allows you to specify fast_rcnn.config options, e.g.
#   --set EXP_DIR seed_rng1701 RNG_SEED 1701

Citing

If you find this repository useful in your research, please consider citing:

@ARTICLE{LR-FTI-FDet,
  author={Y. {Zhang} and M. {Liu} and Y. {Yang} and Y. {Guo} and H. {Zhang}},
  journal={IEEE Transactions on Industrial Informatics}, 
  title={ A Unified Light Framework for Real-time Fault Detection of Freight Train Images}, 
  year={2021},
  volume={},
  number={},
  pages={},
  }

@ARTICLE{Light-FTI-FDet,
  author={Y. {Zhang} and M. {Liu} and Y. {Chen} and H. {Zhang} and Y. {Guo}},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  title={Real-Time Vision-Based System of Fault Detection for Freight Trains}, 
  year={2020},
  volume={69},
  number={7},
  pages={5274-5284},
  }

@INPROCEEDINGS{FTI-FDet,
  author={Y. {Zhang} and K. {Lin} and H. {Zhang} and Y. {Guo} and G. {Sun}},
  booktitle={2018 25th IEEE International Conference on Image Processing (ICIP)}, 
  title={A Unified Framework for Fault Detection of Freight Train Images Under Complex Environment}, 
  year={2018},
  volume={},
  number={},
  pages={1348-1352},
  }

lmomoy/LR-FTI-FDet