Notes for building a DL tool
Deepnotes is an ongoing series of notes for a ML pipeline.
View it on GitHub, Report Errors or make Suggestions
Note: Please kindly note that this legal disclaimer is applied to all the material in this website!
Setting up communications
when we develop machine learning systems for embedded (or remote) targets it is always convenient to have remote access to these machines, It will enable us to run/debug code remotely or even develop at the same time.
Maintaining Reproducibility
Reproducibility of the training variable flow can help debug and improve networks.
Leveraging GStreamer and DeepStream for deep learning pipelines
GStreamer as a media flow backbone to make deep learning modular and DeepStream to get the best out of nvidia’s hardware capabilities.
Using Docker containers
Working with docker containers helped me to fail-recover-improve with less overhead and make things more portable. #using-can-bus-with-nvidia-agx-xavier-devices
Using CAN bus with Nvidia AGX-Xavier
Vehicle inter processor communication is mostly facilitated by the CAN bus. Using AGX-Xavier with the CAN bus opens many doors to create intelligent control systems.
[Nets for Image Processing]
- A PyTorch WalkThrough: Note 1
- Tensors and Data: Note 2
- Building a Network 1
- Backpropagation
- SoftMax loss and Cross Entropy loss gradient: From theory to practice
upcoming..