The dataset consists of 117 images focused on "Dental Radiology Scans," which serve as the foundation for training and evaluating detection and segmentation algorithms in the dental field. This dataset is critical for developing models that can accurately interpret dental images and assist in diagnosis.
To enhance the model's robustness and generalization, 90% of the images were utilized for training. Data augmentation techniques were employed to generate a total of 269 training images, improving the dataset's diversity. The following transformations were applied:
After applying these augmentations, the dataset composition includes:
Out of the total images, 100 images were manually annotated using the COCO JSON Format for object detection. The COCO format is a widely adopted standard for object detection tasks, allowing for efficient integration with various machine learning frameworks, including Detectron2. This meticulous annotation process ensures high-quality training data, critical for achieving robust model performance.
All 269 images and their combined single .json annotation file were utilized for training the model. The training involved using Detectron2, which executed both:
The code saves the Binary Predicted Mask of the test set, providing clear visualizations of the model's performance on unseen data. These binary masks highlight the areas of interest, allowing for quick assessment of model accuracy.
To streamline your object detection workflow, consider automating tasks with Roboflow and MakeSense.ai:
Roboflow: This platform simplifies dataset management by offering tools to preprocess, augment, and version control datasets. By integrating Roboflow, you can significantly enhance model performance and facilitate collaboration.
MakeSense.ai: A powerful annotation tool that allows you to annotate images with bounding boxes, polygons, keypoints, and more. This tool accelerates the data labeling process, enabling you to create high-quality annotations quickly and efficiently, which is essential for training robust models.
To set up the project locally, follow these steps:
Clone the Repository:
git clone https://github.com/arpsn123/Dental-X-RAY-Image-Detection-and-Instance-Segmentation.git cd Dental-X-RAY-Image-Detection-and-Instance-Segmentation
Create a Virtual Environment (recommended):
python -m venv venv source venv/bin/activate # On Windows use `venv\Scripts\activate`
Install Dependencies:
Make sure you have Python 3.7 or later installed. Use the following command to install the required libraries:
pip install -r requirements.txt
Install Detectron2:
Follow the Detectron2 installation instructions to ensure it is properly set up in your environment.
Download the Dataset:
Ensure that you have the dataset stored in the correct directory structure as required by the code.
Run the Training Script:
Execute the training script to begin the training process:
python train.py
Evaluate the Model:
After training, run the evaluation script to assess the model's performance:
python evaluate.py
This project exemplifies the application of cutting-edge machine learning techniques to the field of dental radiology, enhancing the accuracy and efficiency of dental image analysis through advanced object detection and segmentation methodologies.
