KPConvX: Modernizing Kernel Point Convolution with Kernel Attention
This work addresses the need for more efficient and effective point cloud processing in computer vision, representing an incremental advancement over existing kernel point methods.
The authors tackled the problem of improving deep point cloud understanding by modernizing the KPConv architecture, resulting in KPConvX which outperforms current state-of-the-art methods on datasets like ScanObjectNN, Scannetv2, and S3DIS.
In the field of deep point cloud understanding, KPConv is a unique architecture that uses kernel points to locate convolutional weights in space, instead of relying on Multi-Layer Perceptron (MLP) encodings. While it initially achieved success, it has since been surpassed by recent MLP networks that employ updated designs and training strategies. Building upon the kernel point principle, we present two novel designs: KPConvD (depthwise KPConv), a lighter design that enables the use of deeper architectures, and KPConvX, an innovative design that scales the depthwise convolutional weights of KPConvD with kernel attention values. Using KPConvX with a modern architecture and training strategy, we are able to outperform current state-of-the-art approaches on the ScanObjectNN, Scannetv2, and S3DIS datasets. We validate our design choices through ablation studies and release our code and models.