Research map / Detection and segmentation / CNN features
MULLER: Multilayer Laplacian Resizer for Vision
Zhengzhong Tu, Peyman Milanfar, Hossein Talebi
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 |
|---|---|---|---|
| jointly finetune MaxViT-B with MULLER resizer 512->224 | MaxViT-B · ImageNet-1K val | top-1 accuracy: 84.95 -> 85.58 | Improved |
| jointly finetune MaxViT-T with MULLER resizer 512->224 | MaxViT-T · ImageNet-1K val | top-1 accuracy: 83.62 -> 83.95 | Improved |
| add MULLER resizer to ResNet-50 | ResNet-50 · ImageNet-1K val | top-1 accuracy: 75.3 -> 76.2 | Improved |
| add MULLER resizer to MobileNet-v2 | MobileNet-v2 · ImageNet-1K val | top-1 accuracy: 70.5 -> 71.8 | Improved |
| use MULLER instead of prior CNN learned resizer on EfficientNet-B0 | EfficientNet-B0 · ImageNet-1K val | top-1 accuracy: 77.9 -> 78.2 | Improved |
| use MULLER instead of prior CNN learned resizer on ResNet-50 | ResNet-50 · ImageNet-1K val | top-1 accuracy: 76.2 -> 76.2 | No clear effect |
| add MULLER to MaxViT-B for COCO detection at 640x640 | MaxViT-B Mask R-CNN · COCO2017 val | box AP: 51.6 -> 52.3 | Improved |
| add MULLER to MaxViT-T for image aesthetic assessment | MaxViT-T · AVA | PLCC: 0.707 -> 0.729 | Improved |
| vary number of MULLER layers (2, 3, 4, 6) (ablation) | MULLER on MaxViT-B · ImageNet-1K val | top-1 accuracy: 85.58 -> 85.50 | No clear effect |
| vary Gaussian kernel size and std of MULLER (ablation) | MULLER on MaxViT-B · ImageNet-1K val | top-1 accuracy: 85.48 -> 85.58 | No clear effect |
| increase MULLER input resolution from 512 to 768 for 224 output (ablation) | MULLER on MaxViT-B · ImageNet-1K val | top-1 accuracy: 85.58 -> 85.56 | No clear effect |
| reduce MULLER input resolution from 512 to 384 for 224 output (ablation) | MULLER on MaxViT-B · ImageNet-1K val | top-1 accuracy: 85.58 -> 85.45 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MobileNet-v2 — MobileNetV2: Inverted Residuals and Linear Bottlenecks (CVPR 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning To Resize Images for Computer Vision Tasks (ICCV 2021 · similarity 0.89)
- LaplacianFormer:Rethinking Linear Attention with Laplacian Kernel (ICLR 2026 · similarity 0.87)
- Low-Rank Rescaled Vision Transformer Fine-Tuning: A Residual Design Approach (CVPR 2024 · similarity 0.86)
- Win-Win: Training High-Resolution Vision Transformers from Two Windows (ICLR 2024 · similarity 0.84)
- Preserving Deep Representations in One-Shot Pruning: A Hessian-Free Second-Order Optimization Framework (ICLR 2025 · similarity 0.84)
- RandLoRA: Full rank parameter-efficient fine-tuning of large models (ICLR 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.
