TwistNet-2D is a new CNN module that uses directional spiral shifts and normalized channel products to model local second-order texture interactions, outperforming larger backbones on four benchmarks with only 3.5% extra parameters.
Julesz, Textons, the elements of texture perception, and their interactions, Nature 290 (1981) 91–97
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
TwistNet-2D: Learning Second-Order Channel Interactions via Spiral Twisting for Texture Recognition
TwistNet-2D is a new CNN module that uses directional spiral shifts and normalized channel products to model local second-order texture interactions, outperforming larger backbones on four benchmarks with only 3.5% extra parameters.