The Exploration of the Transfer Learning Technique for Globotruncanita genus against the Limited Low-Cost Light Microscope Images
This research study aims to leverage transfer learning-based pre-trained deep learning models to automate the species-level identification of the Globotruncanita genus from low-cost light microscope images. This task is particularly important due to the complex morphological differences across the three species of interest, which carry significant implications for paleoecological interpretations.
Traditionally, the accurate taxa of fossil species has been a labor-intensive process requiring considerable time and expertise, due to the complex structure and morphology of these organisms. The development of an automated system, therefore, represents a significant stride in this research area, particularly given the low-resolution records typically available.
In this study, transfer learning is employed to utilize the strengths of pre-trained deep convolutional neural network (CNN) models. These models have been trained on millions of images and thus have numerous learned filters. This provides a solid foundation for further fine-tuning the model to identify the specific species within the Globotruncanita genus.
To our knowledge, this research study is the first of its kind to apply transfer learning techniques for the species-level identification of the Globotruncanita genus. This represents a significant advancement in the field of paleoenvironmental studies and has potential implications for the future of geological age determination.
The efficacy of this approach will be tested through rigorous analysis and evaluation of model performance. As such, the findings of this study could have significant implications for both the field of paleoenvironmental studies and the broader application of transfer learning in image classification tasks.
This repository contains a collection of Python scripts
It is recommended to use Anaconda Python distribution and Tensorflow with version 2.10.0.