| Summary: | The task of learning from point cloud data is always challenging due to the often occurrence of noise and outliers in the data. Such data inaccuracies can significantly influence the performance of state-of-the-art deep learning networks and their ability to classify or segment objects. While there are some robust deep-learning approaches, they are computationally too expensive for real-time applications. This paper proposes a deep learning solution that includes novel robust pooling layers which greatly enhance network robustness and perform significantly faster than state-of-the-art approaches. The proposed pooling layers replace conventional pooling layers in networks with global pooling operations such as PointNet and DGCNN. The proposed pooling layers look for data mode/cluster using two methods, RANSAC, and histogram, as clusters are indicative of models. We tested the proposed pooling layers on several tasks such as classification, part segmentation, and points normal vector estimation. The results show excellent robustness to high levels of data corruption with less computational requirements as compared to robust state-of-the-art methods. our code can be found at https://github.com/AymanMukh/ModePooling.
|
|---|