The goal of this project is to implement a camera pose estimator based on natural feature tracking, with the RANSAC algorithm to enhance the accuracy of the pose estimation and render a 3d-cube.
In this step, the application detects and describes features from the reference frame. It also identifies the probable matches of these features on the input frame.
To reduce the number of false positives, a feature matching step is performed. Since there may be numerous features with potential false matches, this step aims to refine the matches for better accuracy.
After the feature matching process, the next step is pose estimation. Here, the application assumes that the matched features may contain outliers. To obtain a reliable estimate, the Random Sample Consensus (RANSAC) algorithm is employed. RANSAC helps in finding a robust pose estimate that is not significantly affected by outliers.
Using the estimated pose from the previous step, the application proceeds to render a 3D object. This final step visualizes the object based on the obtained pose information, resulting in a realistic representation.
- Clone the repository:
git clone https://github.com/fuzailpalnak/3dRendering.git - Navigate to the project directory:
cd project_directory - Install the required dependencies:
pip install -r requirements.txt
- Follow data creation steps to render on custom objects
- Run
python rendering/ar.py