note

Depthwise Separable ConvolutionとG検定（ジェネラリスト検定）

「通常の畳み込み（Standard Convolution）を、『空間方向の処理』と『チャンネル方向の処理』の2つのステップに分解することで、計算量とパラメータ数を劇的に削減する手法」のことです。 仕組みの核心: 通常の畳み込みは、「空間的な特徴（縦×横）」と ...
note

【AI基礎】MobileNetのDepthwise Separable Convolution？層毎に大きく畳み込み

MobileNetで使われるDepthwise Separable Convolutionについて、簡単な例を用いて解説します。 Depthwise Separable Convolutionとは？ Depthwise Separable Convolutionは、通常の畳み込み演算を2つのステップに分解することで、計算量を大幅に減らす手法です。 1. 通常の畳み込み（例 ...
IEEE

Depthwise Separable Convolution architectures for 3D Object Recognition *

Abstract: Fast and efficient recognition of 3D objects remains vital in computer vision. View-based recognition has gained increasing attention because it mimics human perception by recognizing ...
IEEE

MixCIM: A Hybrid Computing-in-Memory Macro With Less Data-Movement and Better Memory-Reuse ...

Abstract: Computing-in-memory (CIM) architectures have demonstrated strong potential for edge artificial intelligence (AI) devices due to their enhanced parallelism and energy efficiency. With the ...
GitHub

depthwise_separable_conv_module.py

See https://arxiv.org/pdf/1704.04861.pdf for details. This module can replace a ConvModule with the conv block replaced by two conv block: depthwise conv block and ...
GitHub

Accelerating Vision Foundation Models with Drop-in Depthwise Convolution

Multi-Head Self-Attention in ViTs (left) is approximated by depthwise convolution over the reshaped value tensor (right). We aim to reduce the inference cost of large pre-trained Vision Transformers ...