Natural Scene Segmentation: Deeplabv3+ Pytorch
Budget: €50 – €99 EUR
I am looking to train a Deeplabv3(Plus) model in Pytorch with the objective of segmenting natural scene images.
Key project requirements include:
DATASET is available. There are RGB 256 Images aus Ground Truth data, grayscale heights 256 ,surface_normals images and labels from a function called rgb_to_class {}
The scripts are available:
training.py, trainer.py, custom_dataset.py and model.py
Problem: Pytorch only have a Deeplav3 model to train the custom dataset.
The suitable candidate should demonstrate they can:
* Handle BirdsEyeView Images
* Proficiently use Deeplabv3(Plus) and Pytorch for image segmentation.
I need a few training sequences as well as what has been changed in the parameters and how the result has improved. Some Pictures and datas , tables, whatever.
As an outcome of this project, I expect accurate segmentation of the provided images. The candidate should be able to handle large datasets and work in a timely manner. Previous experience with similar projects is highly appreciated.
Key project requirements include:
DATASET is available. There are RGB 256 Images aus Ground Truth data, grayscale heights 256 ,surface_normals images and labels from a function called rgb_to_class {}
The scripts are available:
training.py, trainer.py, custom_dataset.py and model.py
Problem: Pytorch only have a Deeplav3 model to train the custom dataset.
The suitable candidate should demonstrate they can:
* Handle BirdsEyeView Images
* Proficiently use Deeplabv3(Plus) and Pytorch for image segmentation.
I need a few training sequences as well as what has been changed in the parameters and how the result has improved. Some Pictures and datas , tables, whatever.
As an outcome of this project, I expect accurate segmentation of the provided images. The candidate should be able to handle large datasets and work in a timely manner. Previous experience with similar projects is highly appreciated.