MXNet-1.8.0 & 1.7.0, TensorFlow-2.4.3, 2.3.4 & 1.15.5, PyTorch-1.7.1 & 1.8.1, Neuron, & others. NVIDIA CUDA, cuDNN, NCCL, Intel MKL-DNN, Docker, NVIDIA-Docker & EFA support. For fully managed experience, check: https://aws.amazon.com/sagemaker Root device type: ebs Virtualization type: hvm ENA Enabled: Yes
Run this command, if necessary, to ensure your key is not publicly viewable
chmod 400 ml-test.pem
Connect to your instance using its Public DNS: ec2-34-221-121-230.us-west-2.compute.amazonaws.com
ssh -i "ml-test.pem" ec2-user@ec2-34-221-121-230.us-west-2.compute.amazonaws.com
- Create an conda env for the task:
conda create --name myenvmyenv: the name of your environment. In this case I used 'facialrecog'.
- Ative o ambiente conda:
conda activate facialrecog- Install necessary dependencies (make sure cmake is installed first)::
git clone https://github.com/davisking/dlib.git- Create build folder
cd dlib, mkdir build, cd build- Check compatibility with the machine. Make sure the dlib will not use CUDA and cuDNN. There is a conflict with the architecture and for now I have not been able to find a way to optimize the process. StackOverflow Opened Question
cmake .., cmake --build .- Install setup.py using the C++ lib in python
cd ..
python3 setup.py install- Install jupyter notebook:
conda install jupyter- Install pip in conda env:
conda install pip- Install ipykernel (jupyternotebook):
pip install ipykernel
# preferably do not install only with 'conda install ipykernel'- Install OpenCV:
pip install opencv-pythonOpen the following file in your browser:
face_recognition/flowchart_face_recog.drawio.html
Or see the pipeline pdf Pipeline's FlowChart
/Data contains the /to_detect folder which consists of the images to have faces detected.
They also contain the /to_recog folder where the images have created numpy arrays.
The algorithm is still under construction. Watch this video for a better understanding of how this beta version works:
Resource: