Research map / Architectures and efficiency / CNN design
Learning to Zoom: a Saliency-Based Sampling Layer for Neural Networks
Adria Recasens, Petr Kellnhofer, Simon Stent, Wojciech Matusik, Antonio Torralba
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 |
|---|---|---|---|
| add the saliency sampler layer | AlexNet · GazeCapture (iPad) | distance error (cm): 5.44 -> 3.29 | Improved |
| add the saliency sampler layer | AlexNet · GazeCapture (iPhone) | distance error (cm): 2.63 -> 2.03 | Improved |
| add deformable convolutions instead of the sampler | AlexNet · GazeCapture (iPad) | distance error (cm): 5.44 -> 5.21 | No clear effect |
| add affine STN | AlexNet · GazeCapture (iPad) | distance error (cm): 5.44 -> 4.33 | Improved |
| replace affine STN by TPS STN | AlexNet + STN · GazeCapture (iPad) | distance error (cm): 4.33 -> 4.44 | No clear effect |
| estimate the sampling grid directly without a saliency map | AlexNet · GazeCapture (iPad) | distance error (cm): 5.44 -> 3.91 | Improved |
| add the saliency sampler layer | ResNet-101 · iNaturalist 2017 val | top-1 accuracy (%): 60 -> 65 | Improved |
| add the saliency sampler layer | ResNet-101 · iNaturalist 2017 val | top-5 accuracy (%): 83 -> 86 | Improved |
| add CAM-based crops | ResNet-101 · iNaturalist 2017 val | top-1 accuracy (%): 60 -> 62 | Improved |
| add the saliency sampler layer | Inception V3 · iNaturalist 2017 val | top-1 accuracy (%): 64 -> 66 | Improved |
| add the saliency sampler layer | ResNet-50 · CUB-200 | top-1 accuracy (%): 81.6 -> 84.5 | Improved |
| deepen the saliency network from none to 6 layers (ablation) | RN-101 + Saliency Sampler · iNaturalist 2017 val | top-1 accuracy (%): 60 -> 62 | Improved |
Closest papers by meaning
- Shallow and Deep Convolutional Networks for Saliency Prediction (CVPR 2016 · similarity 0.88)
- There and Back Again: Revisiting Backpropagation Saliency Methods (CVPR 2020 · similarity 0.88)
- Understanding and Visualizing Deep Visual Saliency Models (CVPR 2019 · similarity 0.88)
- A Deeper Look at Saliency: Feature Contrast, Semantics, and Beyond (CVPR 2016 · similarity 0.88)
- Predicting When Saliency Maps Are Accurate and Eye Fixations Consistent (CVPR 2016 · similarity 0.88)
- Saliency Grafting: Innocuous Attribution-Guided Mixup with Calibrated Label Mixing (AAAI 2022 · similarity 0.88)
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.
