Research on the potential application of machine learning in the task of identifying the biological species of a fossil from its image captured by an electron microscope.
- Assessment of the required number of images for training the model.
- Investigation of the impact of the sample's preservation quality in the image on the predictive ability of the created model.
- Determination of the minimum system requirements for the operation of both the developed algorithm and similar neural network-based solutions.
The project was implemented using convolutional neural networks and the functionality of the TensorFlow library. Transfer learning was employed for training the convolutional layers.
