atlas

Research map / Architectures and efficiency / CNN design

TI-Pooling: Transformation-Invariant Pooling for Feature Learning in Convolutional Neural Networks

CVPR 2016 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis · 262 citations (OpenAlex)

Check your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use TI-pooling over 24 rotations instead of a TIRBM baselineMNIST rotation classification · mnist-rot-12kerror (%): 4.2 -> 1.2Helped
Use TI-pooling instead of STN (affine)Half-rotated MNIST · Half-rotated MNISTerror (%): 0.7 -> 0.8Hurt
Use TI-pooling instead of CNN with data augmentationNeuronal segmentation · Neuronal segmentation testpixel error (%): 8.1 -> 7.0Helped
Use TI-pooling instead of MIL over CNNNeuronal segmentation · Neuronal segmentation testpixel error (%): 8.9 -> 7.0Helped
Add dropout (ablation)TI-pooling · Neuronal segmentation testpixel error (%): 7.4 -> 7.0Helped
Increase the number of sampled transformations in the TI-pooling set (ablation)TI-pooling · Neuronal segmentation validationvalidation errorHelped

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy