Research map / Detection and segmentation / CNN features
Semantic Fisher Scores for Task Transfer: Using Objects to Classify Scenes
Mandar Dixit, Yunsheng Li, Nuno M. Vasconcelos
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 |
|---|---|---|---|
| embed semantics in natural parameter space (log) before GMM-FV | SMN-FV · MIT Indoor and SUN | accuracy (%): around 9% | Improved |
| use Gaussian scaling instead of Fisher scaling F^-1/2 (ablation) | DMM-FV · MIT Indoor | accuracy (%): 58.8 -> 67.1 | Improved |
| use Gaussian scaling instead of Fisher scaling F^-1/2 (ablation) | GMM-FV · MIT Indoor | accuracy (%): 57.7 -> 68.5 | Improved |
| change assignment function q to h (ablation) | GMM-FV · MIT Indoor | accuracy (%): 68.7 -> 68.5 | No clear effect |
| use nu(2) pre-softmax embedding instead of nu(1) (ablation) | GMM-FV · MIT Indoor | accuracy (%): 67.7 -> 68.5 | Improved |
| use nu(2) instead of nu(3) (ablation) | GMM-FV · SUN | accuracy (%): 50.47 -> 51.17 | Improved |
| use natural-parameter nu(2) instead of square-root embedding | GMM-FV · MIT Indoor | accuracy (%): 58.95 -> 68.5 | Improved |
| use natural-parameter nu(2) instead of von-Mises embedding | GMM-FV · SUN | accuracy (%): 46.1 -> 51.17 | Improved |
| add log to fc7 features (ablation) | fc7-FV · MIT Indoor | accuracy (%): 65.1 -> 58 | Hurt |
| add log to nu(2) features (ablation) | nu(2)-FV · MIT Indoor | accuracy (%): 68.5 -> 60.97 | Hurt |
| use full-covariance MFA-FS instead of variance GMM-FV | FV pooling of CNN features · MIT Indoor | accuracy (%): 60.7 | Improved |
| random instead of MFA-FS initialization (ablation) | MFAFSNet · MIT Indoor | accuracy (%): 2-3% | Hurt |
Closest papers by meaning
- Object based Scene Representations using Fisher Scores of Local Subspace Projections (NeurIPS 2016 · similarity 0.89)
- Deep Scene Image Classification With the MFAFVNet (ICCV 2017 · similarity 0.87)
- Harnessing Object and Scene Semantics for Large-Scale Video Understanding (CVPR 2016 · similarity 0.86)
- Modality and Component Aware Feature Fusion For RGB-D Scene Classification (CVPR 2016 · similarity 0.86)
- Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction (NeurIPS 2025 · similarity 0.85)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (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.
