Github project containing code written and used for our thesis-work. Includes an active learning evaluation pipeline, some object detection models and data processing.
-
Ensure you have nvidia drivers installed.
-
Follow the detailed install information ("not quickstart") on pip installation homepage.
-
Init conda for shell interaction:
conda init
-
(Linux optional if installation fails) Extend temp folder:
sudo mount -o remount,size=6G /tmp
-
Use pip to install the following packages:
pip install matplotlib pandas tqdm tensorflow_datasets tensorflow_addons python_dotenv
-
ImageNetV2 is currently (05-05-2023) not available through tensorflow_datasets, so it must be downloaded manually. This is hopefully fixed in the future. Download the ImageNetV2 dataset from here