Get Started
Research questionHow can depth compression merge padded convolutions without enlarging kernels for faster embedded inference?Padding complicates analytical layer merging, while enlarged kernels can offset the speed benefits of reducing network depth. The challenge is to merge layers efficiently without increasing kernel size.
Computer Vision
Inference Optimization
Machine Learning
Small / On-device Models
Latest papersRecent research connected to this question, newest first.From Deep to Shallow: Unconstrained and Efficient Layer Merging StrategyThe source evaluates layer merging across multiple architectures and datasets, including inference-speed measurements on real embedded platforms. It does not specify the particular architectures, datasets, or hardware platforms in the abstract.research paper · Sep 4, 2026
Related questions
How can CNN inference fit on bare-metal microcontrollers with severe memory, binary-size, and runtime limits?How can neural networks retain capacity while fitting within fixed parameter and memory budgets?How can we robustly compress LLM KV caches across open-domain inputs without input-specific budget thresholds?How can we compress 3D point-cloud models in ONNX form without source code for constrained edge hardware?
Home
Topics
Search
Library