Counting grapes on images

Contents:

• data Data set of pictures, where each grape has been marked with a coordinate in a respective txt file
• src python source code for model generation, training and result analysis
• remote-results results of the training of the vm
• results results of local training

Main file for training is src/train.py

Concepts of implementation

Each model (subfolders of src/models/) has at least two different training configurations. One for the training with complete pictures in a reduced size and one for split pictures. Each training configuration implements certain methods that are needed for the training methodology (strategy pattern). These methods are injected into the src/Trainer.py class and are used during training. The Trainer class handles all algorithms regarding training, logging and persistence of results/history/model-weights.