A Fast and Open Source Autonomous Perception System.
The more fast your are, the more shadowless you got.
Shadowless is a new generation auto-drive perception system that feel things only in vision(more features maybe add in). We building shadowless on the purpose of establish a fully intelligent and fast drive system.
The 3 main part of shadowless are:
You need to do some setups to run Shadowless, you should download mxnet_ssd model from mxnet official examples repo.
and you should download one video from Youtube or anywhere place it into: videos/
directory. then:
sudo pip3 install -r requirements.txt
Please do 2 steps before you start running python3 main.py
:
Resnet-50 512x512
, directly download url is here, after downloaded untar it into mxnet_ssd/model dir;To run Shadowless after you get all pre-requirements, you can simply do:
python3 main.py
this will start a Shadowless master process to serve camera inputs and do the perception jobs.
So much welcome the community contribute your code to Shadowless, we are now need those features:
Detection with SSD
Detection with FasterRCNN
Detection with RFCN
Lane Segment using OpenCV
Lane Segment with DeepLearning methods
Distance estimate with Objects
Speed estimate with moving objects
Accelerate whole networks to a real-time speed.
this work inspired by Jin Fagang, you should not spread this soft-ware witout any guarantee, please using this under Apache License.