RocNet: Recursive Octree Network for Efficient 3D Deep Representation
This work addresses efficient 3D data compression for applications like shape classification and reconstruction, but it appears incremental as it builds on autoencoder-like networks with recursive octree structures.
The paper tackled the problem of compressing 3D voxel data efficiently by introducing a deep recursive octree network that reduces grids of sizes 32, 64, and 128 to 80 floats in the latent space, achieving maintained accuracy with less memory and shorter training times compared to existing methods.
We introduce a deep recursive octree network for the compression of 3D voxel data. Our network compresses a voxel grid of any size down to a very small latent space in an autoencoder-like network. We show results for compressing 32, 64 and 128 grids down to just 80 floats in the latent space. We demonstrate the effectiveness and efficiency of our proposed method on several publicly available datasets with three experiments: 3D shape classification, 3D shape reconstruction, and shape generation. Experimental results show that our algorithm maintains accuracy while consuming less memory with shorter training times compared to existing methods, especially in 3D reconstruction tasks.