Collection of tasks for fast baseline solving in self-driving problems with deep learning. The offical language support is Pytorch

---

About • Installation • Examples • Inference • Contribution • License

---

Note: This pkg is currently in development. Please open an issue if you find anything that isn't working as expected.

---

What is NNcore

NNcore is a deep learning framework focusing on solving autonomous-driving problems.

• Task-based training and export for multiple framework

Installation

Pip / conda

pip install --upgrade --force-reinstall --no-deps albumentations
pip install qudida
pip install git+https://github.com/HCMUS-ROBOTICS/ssdf-nncore

Other installations

To install nncore and develop locally

git clone https://github.com/HCMUS-ROBOTICS/ssdf-nncore nncore
cd nncore
pip install -e .

---

Examples

We provide some examples here. You can use the interactive version by the colab notebooks belows. For more detail, you can refer to this folder

• Segmentation on LyftDataset
• Segmentation on UITDataset

Perform inference

Export PyTorch checkpoint to ONNX

Use the script provided in examples, then run the below command

python3 torch2onnx.py <checkpoint> --in_shape 1 3 224 224 --inputs input --outputs output

Performing inference

See serve library

Contribution

If you want to contribute to nncore, please follow steps below:

1. Fork your own version from this repository
2. Checkout to another branch, e.g. fix-loss, add-feat.
3. Make changes/Add features/Fix bugs
4. Add test cases in the test folder and run them to make sure they are all passed (see below)
5. Run code format to check formating before making a commit (see below)
6. Push the commit(s) to your own repository
7. Create a pull request

To run tests

pip install pytest
python -m pytest test/

To run code-format

pip install pre-commit
pre-commit install
pre-commit run -a

License

See LICENSE