Skip to content
Tool to evaluate deep-learning detection and segmentation models, and to create datasets
C++ Python CMake C Slice Shell Other
Branch: master
Clone or download
sergiopaniego Merge pull request #152 from JdeRobot/gsoc_2019
Add changes made during GSoC-19
Latest commit 60268b8 Oct 16, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
DeepLearningSuite
datasetGenerator
docs
sampleGenerator/simpleSampleGenerator
samples
.gitignore
.travis.yml
Brewfile
CONTRIBUTING.md
Doxyfile
LICENSE
README.md
install_darknet.sh
install_opencv.sh

README.md

DeepLearningSuite

Build Status

DeepLearning Suite is a set of tool that simplify the evaluation of most common object detection datasets with several object detection neural networks.

The idea is to offer a generic infrastructure to evaluates object detection algorithms againts a dataset and compute most common statistics:

  • Intersecion Over Union
  • Precision
  • Recall

Supported Operating Systems:

  • Linux
  • MacOS
Supported datasets formats:
  • YOLO
  • COCO
  • ImageNet
  • Pascal VOC
  • Jderobot recorder logs
  • Princeton RGB dataset [1]
  • Spinello dataset [2]
Supported object detection frameworks/algorithms
  • YOLO (darknet)
  • TensorFlow
  • Keras
  • Caffe
  • Background substraction
Supported Inputs for Deploying Networks
  • WebCamera/ USB Camera
  • Videos
  • Streams from ROS
  • Streams from ICE
  • JdeRobot Recorder Logs

Sample generation Tool

Sample Generation Tool has been developed in order to simply the process of generation samples for datasets focused on object detection. The tools provides some features to reduce the time on labelling objects as rectangles.

Installation

We have AppImages !!!

Download from here

To run,
First give executable permissions by running
chmod a+x DetectionSuitexxxxx.AppImage
And Run it by
./DetectionSuitexxxxx -c configFile

Though you would need python in your system installed with numpy.
Also, for using TensorFlow, you would need to tensorflow, and similaraly for keras you would need to install Keras.

If required they can be installed by

pip install tensorflow

or

pip install tensorflow-gpu

Similrarly for keras:

pip install keras

Using Docker

First you need to have docker installed in your computer. If no, assuming you are in Ubuntu, install using the following command

 sudo apt install docker.io

Now you need to start the docker deamon using

sudo service docker start

After starting the deamon , get the DetectionSuite docker file using

sudo docker pull jderobot/dl-detectionsuite

Now run the docker image using

sudo docker run -it -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix jderobot/dl-detectionsuite

where the -it flag instructs Docker to allocate a pseudo-TTY connected to the container’s stdin; creating an interactive bash shell in the container ,-e tag is used to set the environment variable, in this case DISPLAY, this is required because you will be using GUI for interaction through out rest of the tutorials. Voila!! You can now use DetectionDuite. Proceed to beginner's tutorial to get started.

Compile and Install from source

To use the latest version of DetectionSuite you need to compile and install it from source. To get started you can either read along or follow these video tutorials.

Requirements

Common deps

Ubuntu MacOS
sudo apt install build-essential git cmake rapidjson-dev libssl-dev
sudo apt install libboost-dev libboost-filesystem-dev libboost-system-dev libboost-program-options-dev
sudo easy_install numpy
brew install cmake boost rapidjson
sudo apt install libgoogle-glog-dev libyaml-cpp-dev qt5-default libqt5svg5-dev
brew install glog yaml-cpp qt

Also, just add qt in your PATH by running:
echo 'export PATH="/usr/local/opt/qt/bin:$PATH"' >> ~/.bash_profile
Install OpenCV 3.4
git clone https://github.com/opencv/opencv.git
cd opencv
git checkout 3.4
mkdir build && cd build
cmake -D WITH_QT=ON -D WITH_GTK=OFF ..
make -j4
sudo make install
brew install opencv

Optional Dependencies

CUDA (For GPU support)

   NVIDIA_GPGKEY_SUM=d1be581509378368edeec8c1eb2958702feedf3bc3d17011adbf24efacce4ab5 && \

     NVIDIA_GPGKEY_FPR=ae09fe4bbd223a84b2ccfce3f60f4b3d7fa2af80 && \
    sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub && \
    sudo apt-key adv --export --no-emit-version -a $NVIDIA_GPGKEY_FPR | tail -n +5 > cudasign.pub && \
     echo "$NVIDIA_GPGKEY_SUM  cudasign.pub" | sha256sum -c --strict - && rm cudasign.pub && \

     sudo sh -c 'echo "deb http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64 /" > /etc/apt/sources.list.d/cuda.list' && \
     sudo sh -c 'echo "deb http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64 /" > /etc/apt/sources.list.d/nvidia-ml.list'

Update and Install

sudo apt-get update
sudo apt-get install -y cuda

Below is a list of Optional Dependencies you may require depending on your Usage.

  • Camera Streaming Support

Detectionsuite can currently read ROS and ICE Camera Streams. So, to enable Streaming support install any one of them.

  • Inferencing FrameWorks

DetectionSuite currently supports many Inferencing FrameWorks namely Darknet, TensorFlow, Keras and Caffe. Each one of them has some Dependencies, and are mentioned below.

Choose your Favourite one and go ahead.

  • Darknet (jderobot fork)

    Darknet supports both GPU and CPU builds, and GPU build is enabled by default. If your Computer doesn't have a NVIDIA Graphics card, then it is necessary to turn of GPU build in cmake by passing -DUSE_GPU=OFF as an option in cmake.
    git clone https://github.com/JdeRobot/darknet
    cd darknet
    mkdir build && cd build

For GPU users:

cmake -DCMAKE_INSTALL_PREFIX=<DARKNET_DIR> ..

For Non-GPU users (CPU build):

cmake -DCMAKE_INSTALL_PREFIX=<DARKNET_DIR> -DUSE_GPU=OFF ..

Change <DARKNET_DIR> to your custom installation path.

make -j4
sudo make -j4 install

  • TensorFlow

Only depedency for using TensorFlow as an Inferencing framework is TensorFlow. So, just install TensorFlow. Though it should be 1.4.1 or greater.

  • Keras

Similarly, only dependency for using Keras as an Inferencing is Keras only.

  • Caffe

For using Caffe as an inferencing framework, it is necessary to install OpenCV 3.4 or greater.

Note: Be Sure to checkout the Wiki Pages for tutorials on how to use the above mentioned functionalities and frameworks.

How to compile DL_DetectionSuite:

Once you have all the required Dependencies installed just:

    git clone https://github.com/JdeRobot/DetectionSuite
    cd DetectionSuite/DeepLearningSuite
    mkdir build && cd build
    cmake ..

To enable Darknet support with GPU:

    cmake -DARKNET_PATH=<DARKNET_INSTALLETION_DIR> -DUSE_GPU_DARKNET=ON ..

Note: GPU support is enabled by default for other Frameworks

    make -j4

NOTE: To enable Darknet support just pass an optinal parameter in cmake -D DARKNET_PATH equal to Darknet installation directory, and is same as <DARKNET_DIR> passed above in darknet installation.

Once it is build you will find various executables in different folders ready to be executed 😄.

Starting with DetectionSuite

The best way to start is with our beginner's tutorial for DetectionSuite. If you have any issue feel free to drop a mail vinay04sharma@icloud.com or create an issue for the same.

Sample of input and output

Screenshot

Screenshot

Screenshot

You can’t perform that action at this time.