Welcome to the repository dedicated to our Siamese Neural Network (SNN) project for diagnosing endoscopy videos and images. Here, you will find all pertinent information, including experiments, examples, dataset links, prototype, and server source code. Below is an organized overview of the contents available for easy navigation. For detailed instructions on running the code, please refer to the README file provided in each respective folder.
In this folder, you will find essential datasets utilized in our experiments. The README file within this directory contains links to the following datasets:
- Old SNN Training Dataset: Dataset used for training the previous version of our Siamese Neural Network.
- Base Network Training Dataset: Dataset utilized for training the base network.
- SNN Training Dataset: Dataset specifically curated for training the Siamese Neural Network.
As these datasets are too large for direct upload to Github, kindly download them from the provided links. After downloading, ensure to adjust the configuration file paths to correctly reference the appropriate dataset. If you are running the example, make necessary path modifications within the provided IPYNB notebook file.
Contained within this folder is an illustrative IPython notebook demonstrating the implementation of our Siamese Neural Network. Detailed explanations of the process and insights into interpreting the results are provided for clarity.
This folder encompasses all source code utilized in the creation, training, and testing of various networks. Included are implementations for:
- Bounding Box Network
- Segmentation Mask Network
- Old SNN Network
- Proper Base Network
- Proper SNN
Explore this directory for comprehensive insights into our experimental methodologies.
Here lies the source code for both the frontend and server components of our prototype. This folder contains crucial elements for deploying and interacting with the prototype system.
Feel free to navigate through each folder for an in-depth understanding of our project components. If you have any inquiries or require further assistance, do not hesitate to reach out. We appreciate your interest in our work!