Get Started
Home
Topics
Search
Library
Research questionHow can we compress 3D point-cloud models in ONNX form without source code for constrained edge hardware?Edge devices have limited compute and memory, while vendor-distributed 3D point-cloud models may be available only as ONNX binaries. Compression tools that depend on the original source code cannot directly operate under that access constraint.
AI
Computer Vision
Inference Optimization
Machine Learning
Research Paper
Small / On-device Models
Technology
Latest papersRecent research connected to this question, newest first.H3DNAS: Hardware-Aware ONNX-Native 3D Point Cloud Model CompressionThe work concerns ONNX computational graphs for 3D point-cloud models targeting edge hardware such as the NVIDIA Jetson Orin Nano. Reported evidence covers PointNet, PointNet++, and PointMLP on ModelNet40, measuring parameter reduction, inference speed, and accuracy; the search does not require original source code, an architecture class definition, or gradient access.research paper · Sep 2, 2026
Related questions
How can image-to-shape diffusion transformers be compressed for resource-constrained 3D generation without losing geometric fidelity?How can existing 3D Gaussian Splatting models be compacted for storage and transmission while preserving novel-view fidelity?How can depth compression merge padded convolutions without enlarging kernels for faster embedded inference?How can novel-view synthesis represent continuous 3D scenes with less capacity and optimization than NeRFs?