Research map / Label-efficient and robust learning / Recognition and long-tail
A Discriminative View of MRF Pre-Processing Algorithms
Chen Wang, Charles Herrmann, Ramin Zabih
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 |
|---|---|---|---|
| use discriminative rule instead of DEE pre-processing | discriminative pre-processing + alpha-expansion · Denoising-sq | speedup: 1.20x -> 11.83x | Improved |
| use discriminative rule instead of DEE pre-processing | discriminative pre-processing + alpha-expansion · Inpainting | labeled variables: 21.05% -> 74.29% | Improved |
| use discriminative rule instead of DEE pre-processing | discriminative pre-processing + alpha-expansion · Stereo | energy change: 0.00% -> -0.06% | Improved |
| use discriminative rule instead of IRI pre-processing | discriminative pre-processing + alpha-expansion · Color-seg-n4 | speedup: 3.67x -> 7.02x | Improved |
| use discriminative rule instead of IRI pre-processing | discriminative pre-processing + alpha-expansion · Color-seg-n8 | energy change: -0.10% -> +0.04% | Hurt |
| reject expansion moves that increase energy (ablation) | discriminative pre-processing + alpha-expansion · Challenging datasets | final energy | Hurt |
| raise kappa from 0.7 to 1.0 (ablation) | discriminative pre-processing + alpha-expansion · Stereo | precision: 90.40% -> 100.00% | Improved |
| raise kappa from 0.7 to 1.0 (ablation) | discriminative pre-processing + alpha-expansion · Stereo | recall: 91.31% -> 9.26% | Hurt |
| raise kappa from 0.7 to 0.8 (ablation) | discriminative pre-processing + alpha-expansion · Inpainting | precision: 95.11% -> 99.88% | Improved |
| raise kappa from 0.7 to 0.8 (ablation) | discriminative pre-processing + alpha-expansion · Inpainting | recall: 90.51% -> 47.06% | Hurt |
| raise kappa from 0.7 to 1.0 (ablation) | discriminative pre-processing + alpha-expansion · Denoising-ts | recall: 98.61% -> 94.62% | No clear effect |
| add worst-case bound rule delta<=epsilon at kappa>=0.8 (ablation) | discriminative pre-processing + alpha-expansion · speedup and energy | speedup and energy | No clear effect |
Closest papers by meaning
- Relaxation-Based Preprocessing Techniques for Markov Random Field Inference (CVPR 2016 · similarity 0.90)
- Efficient semidefinite-programming-based inference for binary and multi-class MRFs (NeurIPS 2020 · similarity 0.88)
- Bayesian Exploration of Pre-trained Models for Low-shot Image Classification (CVPR 2024 · similarity 0.85)
- Neural Markov Random Field for Stereo Matching (CVPR 2024 · similarity 0.85)
- Revisiting Semi-Supervised Learning in the Era of Foundation Models (NeurIPS 2025 · similarity 0.84)
- Discriminative Score Function: Turning Pretrained Models into Functional Generative Priors (NeurIPS 2026 · 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.
