A collection of the data created during our ML4B project.
The "extracted" directory contains 1000 images of grapes processed by us from the Embrapa Wine Grape Instance Segmentation Dataset. The exact steps for processing can be found in this Google Colab notebook. Please note that the actual processing generated over 4000 images, but GitHub enforces a limit of 1000 images maximum per directory. The full collection with all generated images is available in the extracted.zip file.
The "generated" directory contains 1000 images of grapes that we generated using a neural network and the StyleGAN2 program from NVlabs. We trained the neural network with images from the "extracted" directory. Our trained network is too large to upload to GitHub, but can be downloaded here. It is a .pkl file ready to be used with StyleGAN2.
The images in /extracted/ and extracted.zip are a derivative of "Embrapa WGISD" by Embrapa Agricultural Informatics, maintained by Thiago Santos used under Creative Commons BY-NC 4.0 (Attribution-NonCommercial 4.0 International license). These images are licensed under Creative Commons BY-NC 4.0 (Attribution-NonCommercial 4.0 International license) by Albin Lokaj and Ferdinand Eckl. (Attribution according to best practices)
The images in /generated/ are licensed under CC BY 4.0 by Albin Lokaj and Ferdinand Eckl.