Pith. sign in

REVIEW 5 cited by

Making Convolutional Networks Shift-Invariant Again

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1904.11486 v2 pith:LJOFIZT7 submitted 2019-04-25 cs.CV cs.LG

Richard Zhang This is my paper
classification cs.CVcs.LG
keywords networksconvolutionaldeepdownsamplinginputmax-poolingmodernobserve
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Modern convolutional networks are not shift-invariant, as small input shifts or translations can cause drastic changes in the output. Commonly used downsampling methods, such as max-pooling, strided-convolution, and average-pooling, ignore the sampling theorem. The well-known signal processing fix is anti-aliasing by low-pass filtering before downsampling. However, simply inserting this module into deep networks degrades performance; as a result, it is seldomly used today. We show that when integrated correctly, it is compatible with existing architectural components, such as max-pooling and strided-convolution. We observe \textit{increased accuracy} in ImageNet classification, across several commonly-used architectures, such as ResNet, DenseNet, and MobileNet, indicating effective regularization. Furthermore, we observe \textit{better generalization}, in terms of stability and robustness to input corruptions. Our results demonstrate that this classical signal processing technique has been undeservingly overlooked in modern deep networks. Code and anti-aliased versions of popular networks are available at https://richzhang.github.io/antialiased-cnns/ .

Discussion (0). Sign in to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. From Spatial to Spectral: An Efficient, Frequency-Guided Feature Representation Learner for Small Object Detection

    cs.CV 2026-06 unverdicted novelty 6.0 of 10

    Proposes DERNet with Decompose-Enhance-Reconstruct operator and three plug-and-play modules to shift small object detection from spatial to spectral feature processing, claiming better performance than YOLOv11 with 1/...

  2. Parameter-Efficient Architectural Modifications for Translation-Invariant CNNs

    cs.CV 2026-04 unverdicted novelty 5.0 of 10

    Strategic insertion of Global Average Pooling layers in VGG-16 reduces trainable parameters by 98%, maintains 66.4% ImageNet Top-1 accuracy, doubles translation robustness, and yields superior Spearman correlations in...

  3. GeomPrompt: Geometric Prompt Learning for RGB-D Semantic Segmentation Under Missing and Degraded Depth

    cs.CV 2026-04 unverdicted novelty 5.0 of 10

    GeomPrompt learns a task-driven geometric prompt from RGB alone to substitute for missing or degraded depth in frozen RGB-D semantic segmentation models, yielding up to +6.1 mIoU gains on SUN RGB-D while being faster ...

  4. Geometrically Constrained and Token-Based Probabilistic Spatial Transformers

    cs.CV 2025-09 conditional novelty 5.0 of 10

    A probabilistic, component-wise spatial transformer using a shared frozen tokenizer and augmentation-guided alignment loss improves geometric robustness in fine-grained moth classification.

  5. A Spatial-Frequency Aware Multi-Scale Fusion Network for Real-Time Deepfake Detection

    cs.CV 2025-08 conditional novelty 5.0 of 10

    SFMFNet uses wavelet-frequency gating, token-selective cross-attention, and blur pooling to reach 0.8682 average cross-dataset AUC with only 1.27 GFLOPs and 6.64M parameters.

Pith tools