Research map / Architectures and efficiency
CNN design: research map
853 accepted papers on CNN design in Architectures and efficiency, from ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026), grouped into 3 clusters and 13 approaches. The busiest year so far is 2019.
Within Architectures and efficiency, its share shrank from 7.0% in 2023–24 to 2.9% in 2025–26 (137 → 102 papers at ICML, NeurIPS, CVPR and AAAI, the venues with data for all four years).
Explore CNN design in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
Approaches and key papers
cnn · filter · trained · 395 papers
Approaches in this cluster:
- Filter design and trained-filter analysis (128 papers)
Redesigns CNN filters through shaping, structured filter banks, and empirical analysis of learned filters. - Compact CNN architectures (114 papers)
Condenses and refines convolutional networks via binary structure, learned shapes, and PDE views. - Theory of convolutional inductive bias (64 papers)
Analyzes invariance, stability, and Lipschitz properties of convolutional layers and pooling geometry. - CNN filter pruning (43 papers)
Prunes or factorizes convolutional filters using frequency-domain, gating, and redundancy criteria. - Biologically inspired CNN explanation (46 papers)
Explains CNN behavior and links networks to human visual cortex and perceptual learning.
Most cited and most cited since 2024:
- Densely Connected Convolutional Networks (CVPR 2017 · 45,904 citations)
- An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale (ICLR 2021 · 21,812 citations)
- RepViT: Revisiting Mobile CNN From ViT Perspective (CVPR 2024 · 30 citations)
- Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods (CVPR 2024 · 17 citations)
convolution · kernel · spatial · 344 papers
Approaches in this cluster:
- Efficient CNN scaling and dynamic inference (104 papers)
Scales and compresses CNNs with compound scaling, dynamic spatial computation, and channel-wise convolutions. - Dense matching and estimation CNNs (93 papers)
Designs convolutional descriptors and fusion networks for matching, optical flow, and stereo confidence. - Invertible and wide-network analysis (65 papers)
Builds invertible convolutional networks and analyzes infinitely wide networks through kernel theory. - Large and dynamic kernel design (62 papers)
Rethinks convolution kernel size, shape, and dynamic composition to enlarge receptive fields. - Higher-order pooling (20 papers)
Replaces average pooling with kernel, wavelet, stochastic, and bilinear pooling in CNNs.
Most cited and most cited since 2024:
- Rethinking the Inception Architecture for Computer Vision (CVPR 2016 · 31,549 citations)
- Squeeze-and-Excitation Networks (CVPR 2018 · 30,472 citations)
- EMCAD: Efficient Multi-scale Convolutional Attention Decoding for Medical Image Segmentation (CVPR 2024 · 506 citations)
- InceptionNeXt: When Inception Meets ConvNeXt (CVPR 2024 · 401 citations)
equivariant · equivariance · group · 114 papers
Approaches in this cluster:
- Group symmetry equivariant networks (55 papers)
Builds networks equivariant to Lie groups, diffeomorphisms, and gauge symmetries. - Steerable rotation-equivariant CNNs (40 papers)
Designs steerable filters for rotation, scale, and Euclidean equivariance in convolutional networks. - Group convolution theory (19 papers)
Develops theory and efficient designs of group-equivariant convolutional networks on homogeneous spaces.
Most cited and most cited since 2024:
- Harmonic Networks: Deep Translation and Rotation Equivariance (CVPR 2017 · 594 citations)
- Group Equivariant Convolutional Networks (ICML 2016 · 552 citations)
- Making Vision Transformers Truly Shift-Equivariant (CVPR 2024 · 9 citations)
- Rotation-Equivariant Self-Supervised Method in Image Denoising (CVPR 2025 · 5 citations)
Related topics in Architectures and efficiency
- Pruning and efficient tuning (1,147)
- Attention and transformers (2,296)
- Neural architecture search (380)
- Quantization (646)
- RNNs and normalization (2,827)
- Spiking networks (289)
- Mixture of experts (300)