AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Convolutional Neural Networks Are Not Invariant to Translation, but They Can Learn to Be

Valerio Biscione, Jeffrey S. Bowers

JMLR 2021 · Architectures and efficiency › CNN design · approach: Steerable rotation-equivariant CNNs · 25 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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
pretrain on ImageNet before fine-tuning on one locationCNNs (ResNet-50, VGG16, VGG16bn, AlexNet, DenseNet-121) · one-location datasets, fully-translated testaccuracyImproved
pretrain on ImageNetVGG16 · EMNISTaccuracy (%): 5±0.76 -> 84±3.12Improved
add global average pooling to untrained VGG16/VGG16bn (ablation)VGG16, VGG16bn · one-location datasetsaccuracy (%): ~98Improved
global average pooling already present in ResNet-50 (ablation)untrained ResNet-50 · one-location datasetsaccuracyNo clear effect
add batch normalization (ablation)untrained VGG16 · one-location datasetsaccuracyImproved
add batch normalization (ablation)ImageNet-pretrained VGG16 · one-location datasets (all but LEEK10)accuracyHurt
add anti-aliased max pooling or fully-convolutional padding (ablation)VGG16bn · KMNIST, MNIST, other one-location datasetsaccuracyMixed
pretrain on fully-translated dataset instead of one-location fine-tune onlyResNet-50, VGG16, VGG16bn, AlexNet · Invariance Metric on novel datasetsInvariance Metric: < 0.2 -> 0.83±0.1Improved
fine-tune on one-location data after translated pretraining on same datasetCNN pretrained on fully-translated ShapeNet · ShapeNet fully-translated testaccuracy (%): at least 95 -> 79.56Hurt
pretrain on a more complex translated dataset than the fine-tuning one (ablation)CNNs · cross-dataset fine-tuningaccuracyImproved
train on whole canvas with classes segregated by area (no full translation) (ablation)VGG16 · EMNIST18 / MNISTaccuracyNo clear effect
leave upper-right quadrant empty during pretraining (ablation)VGG16 · EMNISTaccuracy on untrained areaMixed

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