by Huang Yuyao, Wen YongKun
- A differentiable Perspective Transformer Layer inspired by Spatial Transformer Networks. See
bevutils/layers/perspective_transformer.py
andsamples/basic/sample.py
. - A simulated dataset. See
bevutils/datasets/simulated.py
andsamples/simulated_dataset/demo_dataset.py
. - A training system inspired by victoresque's template and mmdetection
pip install git+https://github.com/huangyuyao/bevutils.git@master
git clone https://github.com/huangyuyao/bevutils.git
cd bevutils
pip install -e .
You can make use of functions, datasets and layers of this project by import the functionalities from anywhere.
If you want to use the capability of training a new network, you can do three things:
- registry your custom network modules(TBD),
- write a config file,
- and call `bevtrain -c config.json` or `bevtest -c config.json` tools to reuse the training process. (TBD)
You can refer to samples/perspective_transformer_network
as a quick tutorial. (TBD)
This project is licensed under the MIT License. See LICENSE for more details