Home

Videoanalytics is a library for exploring ways of extracting information from video.

• Official documentation (readthedocs)

Instructions for developers

Running in conda environment with GPU

A GPU conda environment is provided for development.

Import conda environment (GPU):

conda env create -f videoanalytics-gpu.yml

Some examples are provided as jupyter notebooks.

conda activate videoanalytics-gpu.yml
jupyter notebook .

Posible workaround for error:

2021-07-04 10:13:53.920896: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library 'libcudart.so.10.1'; dlerror: libcudart.so.10.1: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/nhorro/catkin_ws/devel/lib:/opt/ros/melodic/lib
2021-07-04 10:13:53.920969: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.

cd /usr/local/cuda/lib64/
sudo ln -sf libcudart.so.10.2 libcudart.so.10.1

Documentation

Generate HTML documentation.

make html

Generate PDF documentation.

make latexpdf

Update/generate UML diagrams (requires PlantUML)

make uml

Update github pages documentation.

make githubpages

Training a custom object detector with DarkNet

FIXME

Exporting YOLOv4 weights to Tensorflow

Follow instructions from tensorflow-yolov4-tflite.

Training a custom feature extractor for DeepSORT

FIXME