Research map / Architectures and efficiency / CNN design
CNN^{2}: Viewpoint Generalization via a Binocular Vision
Wei-Da Chen, Shan-Hung (Brandon) Wu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| CNN2 binocular architecture instead of vanilla CNN | CNN · ModelNet2D unseen viewpoints | accuracy: 0.907 -> 0.941 | Improved |
| CNN2 instead of vanilla CNN | CNN · SmallNORB unseen viewpoints | accuracy: 0.722 -> 0.865 | Improved |
| CNN2 instead of vanilla CNN | CNN · RGB-D Object unseen viewpoints | accuracy: 0.795 -> 0.868 | Improved |
| CNN2 instead of CapsuleNet for training speed | CNN2 · SmallNORB | training time (min): 1328 -> 121 | Improved |
| replace CNN2 blocks with Big-Little modules (CNN2+BL) (ablation) | CNN2 · ModelNet2D | accuracy: 0.941 -> 0.918 | Hurt |
| replace CNN2 blocks with Big-Little modules (CNN2+BL) (ablation) | CNN2 · RGB-D Object | accuracy: 0.868 -> 0.778 | Hurt |
| tie weights of dual feedforward pathways (single pathway) (ablation) | CNN2 · 3D viewpoint generalization | accuracy | Hurt |
| drop parallax augmentation (ablation) | CNN2 · 3D viewpoint generalization | accuracy | Hurt |
| replace CM pooling with max pooling (ablation) | CNN2 · 3D viewpoint generalization | accuracy | Hurt |
| add CM pooling to vanilla CNN | CNN · RGB-D Object | accuracy | Improved |
| pool before convolution instead of after (ablation) | CNN2 · 3D viewpoint generalization | accuracy | Hurt |
| early or late fusion only instead of fusion at all layers (ablation) | CNN2 · RGB-D Object | accuracy | Hurt |
Closest papers by meaning
- Volumetric and Multi-View CNNs for Object Classification on 3D Data (CVPR 2016 · similarity 0.87)
- MVS2D: Efficient Multi-View Stereo via Attention-Driven 2D Convolutions (CVPR 2022 · similarity 0.86)
- Learning from Unlabelled Videos Using Contrastive Predictive Neural 3D Mapping (ICLR 2020 · similarity 0.86)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.85)
- The 3D-PC: a benchmark for visual perspective taking in humans and machines (ICLR 2025 · similarity 0.85)
- Beyond Human Perception: Understanding Multi-Object World from Monocular View (CVPR 2025 · similarity 0.84)
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.
