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
- Linux
- MacOS
- YOLO
- COCO
- ImageNet
- Pascal VOC
- Jderobot recorder logs
- Princeton RGB dataset [1]
- Spinello dataset [2]
- YOLO (darknet)
- TensorFlow
- Keras
- Caffe
- Background substraction
- WebCamera/ USB Camera
- Videos
- Streams from ROS
- Streams from ICE
- JdeRobot Recorder Logs
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.
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
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 dl-detectionsuite docker file using
sudo docker pull jderobot/dl-detectionsuite
Now run the docker image using
sudo docker run -it -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix jderobot/dl-detectionsuite
where the -it
flag instructs Docker to allocate a pseudo-TTY connected to the container’s stdin; creating an interactive bash
shell in the container ,-e
tag is used to set the environment variable, in this case DISPLAY
, this is required because you will be using GUI for interaction through out rest of the tutorials.
Voila!! You can now use dl-detectionsuite. Proceed to beginner's tutorial to get started.
To use the latest version of dl-DetectionSuite you need to compile and install it from source. To get started you can either read along or follow these video tutorials.
Ubuntu | MacOS |
sudo apt install build-essential git cmake rapidjson-dev libssl-dev sudo apt install libboost-dev libboost-filesystem-dev libboost-system-dev libboost-program-options-dev |
sudo easy_install numpy brew install cmake boost rapidjson |
sudo apt install libgoogle-glog-dev libyaml-cpp-dev qt5-default libqt5svg5-dev |
brew install glog yaml-cpp qt Also, just add qt in your PATH by running: echo 'export PATH="/usr/local/opt/qt/bin:$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 |
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.
Detectionsuite can currently read ROS and ICE Camera Streams. So, to enable Streaming support install any one of them.
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 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
Only depedency for using TensorFlow as an Inferencing framework is TensorFlow. So, just install TensorFlow. Though it should be 1.4.1 or greater.
Similarly, only dependency for using Keras as an Inferencing is Keras only.
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.
Once you have all the required Dependencies installed just:
git clone https://github.com/JdeRobot/dl-DetectionSuite
cd dl-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 😄.
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.