Jacinto AI DevKit

This repository provides various training tools for AI, including Deep Learning, traditional Machine Learning and Computer Vision. This repository does not contain any code, but has submodule links that point to the code.

Embedded AI, especially Embedded Inference of Deep Learning models is quite challenging - due to high compute requirements. We show several low complexity Deep Learning models that make Inference on Low Power Embedded Platforms practical. We also have examples of Calibration/Training to make models Quantization friendly.

Several of these models have been verified to work on TI's Jacinto Automotive Processors - but they are also generic and can be used on a variety of Embedded Platforms. This code is primarily intended for learning and research.

PyTorch based training of Object Detection, Semantic Segmentation (and a variety of similar Pixel2Pixel tasks), Multi-Task Learning, Calibration for Quantization, Quantization Aware Training.
TensorFlow2.0 based training and quantization examples (coming soon...)
Caffe based training of Object Detection, Semantic Segmentation.
HOG/Adaboost based training for Object Detection - i.e. using traditional ML.

Obtaining the code

After cloning this repository, execute the following command to pull the submodules.

./git_submodule_update.sh

If there is any issue with the above script, you can visit https://git.ti.com/jacinto-ai-devkit and get the URLs to manually clone/pull the respositories.

Submodules

pytorch-jacinto-ai-devkit: PyTorch (deep learning framework) based training scripts, quantization tools.
pytorch-mmdetection: PyTorch (deep learning framework) based Object Detection training
caffe-jacinto and caffe-jacinto-mdoels: caffe based training tools for sparse CNN models.
acf-jacinto - training tool for HOG/ACF/AdaBoost Object Detector (traditional machine learning based)

Issue Tracker:

While you can file issues using the issues tab in this repository, e2e may be more closely monitored and tracked: https://e2e.ti.com/support/j721e/f/1026/tags/jacinto_2D00_ai_2D00_devkit.
If you are filing an issue via e2e, kindly include jacinto-ai-devkit in the tags (at the end of the page as you create a new issue) so that we get notified quickly.

Notes:

Each of those submodules have detailed documentation and separate LICENSE files. Click on the links to open the documentation.
If you click on one of the links above - it will navigate to a repository hosted in https://git.ti.com/jacinto-ai-devkit. From there, you can click on one of the tabs to get more information on that repository.
For example, the about tab shows documentation. The summary tab lists all the branches, commit information and links for cloning that repository (i.e. if you would like to directly clone without using the above submodule command).

License

Please see the LICENSE file for more information about the license under which this repository is made available. The LICENSE file of each submodule is inside that submodule.

bobdeng1974 / jacinto-ai-devkit Goto Github PK