Example notebooks on building PyTorch, preparing data and training as well as an updated project from a PyTorch MaskRCNN port
Inspired by: https://github.com/multimodallearning/pytorch-mask-rcnn
NOTES:
System Tested (Linux and NVIDIA GPU required with CUDA/cuDNN):
These sets of Jupyter notebooks may also be run in a Docker container running on CUDA-capable GPU hardware.
Work through the notebooks:
Adult Schoolmaster Snappers (Lutjanus apodus); Source: Florent Charpin, http://reefguide.org/pixhtml/schoolmaster2.html
DataSet
class (fish_pytorch_style.py
)If you wish to build PyTorch latest or from a commit, follow one of the two notebooks:
TIP: You can run this project inside a Docker image such as the
rheartpython/cvdeep
public image that has many Deep Learning frameworks preinstalled. (more info at https://github.com/michhar/custom-jupyterhub-linux-vm)