Research map / Vision-language and multimodal / Human and embodied modeling
Write a Classifier: Predicting Visual Classifiers from Unstructured Text
Mohamed Elhoseiny, Ahmed Elgammal, Babak Saleh
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 seen-example negative-side constraints (ablation) | linear formulation E · Oxford Flowers (one fold) | avg AUC: 0.59 -> 0.65 | Improved |
| add constraint t*^T W c >= l (ablation) | linear formulation E · Oxford Flowers (one fold) | avg AUC: 0.58 -> 0.65 | Improved |
| combine regression (TGP) and domain transfer with constraints | linear classifier prediction · Oxford Flowers | avg AUC: 0.62 -> 0.68 | Improved |
| combine regression (TGP) and domain transfer with constraints | linear classifier prediction · CUB Birds | avg AUC: 0.61 -> 0.62 | Improved |
| add constraints to GPR | GPR · Oxford Flowers | avg AUC: 0.54 -> 0.62 | Improved |
| add constraints to TGP | TGP · Oxford Flowers | avg AUC: 0.58 -> 0.63 | Improved |
| add constraints to domain transfer (CDT) | DT · Oxford Flowers | avg AUC: 0.62 -> 0.64 | Improved |
| use constrained DT alone (gamma=0, no TGP term) (ablation) | formulation E · Oxford Flowers | avg AUC: 0.64 -> 0.68 | Hurt |
| predict kernel classifier with one-class SVM adjusted DT instead of linear classifier | DT-based prediction · Flower | |Nsc| to |Nsc+1| recall (%): 31.33 -> 40.34 | Improved |
| predict kernel classifier with one-class SVM adjusted DT instead of linear classifier | DT-based prediction · CUB Birds | |Nsc| to |Nsc+1| recall (%): 36.56 -> 44.05 | Improved |
| replace linear classifier with SVM-DT kernel | classifier prediction · Flower | MAU (%): 5.93 -> 9.1 | Improved |
| replace linear classifier with SVM-DT kernel | classifier prediction · CUB Birds (MKL image kernel) | MAU (%): 2.74 -> 4.10 | Improved |
Closest papers by meaning
- Discovering Unwritten Visual Classifiers with Large Language Models (ECCV 2024 · similarity 0.89)
- Visual Classification via Description from Large Language Models (ICLR 2023 · similarity 0.88)
- Online Collaborative Learning for Open-Vocabulary Visual Classifiers (CVPR 2016 · similarity 0.87)
- Language-Informed Visual Concept Learning (ICLR 2024 · similarity 0.86)
- ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.86)
- Why are Visually-Grounded Language Models Bad at Image Classification? (NeurIPS 2024 · similarity 0.86)
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.
