Code to prepare datasets and do model training and evaluation, in the scope of this research project:
Deep-Learning Image Segmentation - Towards Tea Leaves Harvesting by Automomous Machine
- Python 3.x
- Tensorflow >=1.8 (tested with 1.8 and 1.9)
python3 src/run.py
or by calling predefined tasks in root directory.
- bake: create tf-record dataset from picture/mask/weight images
- preview: read and display content of a tf-record dataset
- train: train a single model with given parameters
- trainmulti: train multiple models
- eval: evaluate a model
- predict: report and display model performance on unseen data
Example of model topologies in YAML format are located in topologies sub-folder.
Example of bake and training parameters for single and multiple session are located in config sub-folder.
GPLv3