AI Frontier Lectures
AI Frontier Lectures
Sep 9, 2025 · Artificial Intelligence

Can UniConvNet Expand Receptive Fields While Preserving Gaussian Distribution?

The paper introduces UniConvNet, a novel convolutional architecture that expands the effective receptive field (ERF) of ConvNets without breaking the asymptotically Gaussian distribution (AGD), achieving superior accuracy‑parameter and accuracy‑FLOPs trade‑offs across image classification, detection, and segmentation benchmarks.

Effective Receptive FieldUniConvNetconvolutional neural networks
0 likes · 9 min read
Can UniConvNet Expand Receptive Fields While Preserving Gaussian Distribution?