In Defense of Classical Image Processing: Fast Depth Completion on the CPU
This work addresses depth completion for autonomous driving and robotics by demonstrating that classical methods can surpass data-driven approaches, offering a data-independent and efficient solution.
The paper tackles depth completion of sparse LIDAR data by proposing a fast, classical image processing algorithm that runs on the CPU, outperforming neural network methods on the KITTI benchmark and ranking first at submission.
With the rise of data driven deep neural networks as a realization of universal function approximators, most research on computer vision problems has moved away from hand crafted classical image processing algorithms. This paper shows that with a well designed algorithm, we are capable of outperforming neural network based methods on the task of depth completion. The proposed algorithm is simple and fast, runs on the CPU, and relies only on basic image processing operations to perform depth completion of sparse LIDAR depth data. We evaluate our algorithm on the challenging KITTI depth completion benchmark, and at the time of submission, our method ranks first on the KITTI test server among all published methods. Furthermore, our algorithm is data independent, requiring no training data to perform the task at hand. The code written in Python will be made publicly available at https://github.com/kujason/ip_basic.