Research map / Detection and segmentation / CNN features
Energy-Based Models for Deep Probabilistic Regression
Fredrik K. Gustafsson, Martin Danelljan, Goutam Bhat, Thomas B. Schön
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 |
|---|---|---|---|
| replace direct regression with energy-based model | Faster-RCNN box regression · COCO 2017 test-dev | AP (%): 37.2 -> 39.4 | Improved |
| replace confidence (IoU) regression with energy-based NLL training | IoU-Net · COCO 2017 test-dev | AP (%): 38.3 -> 39.4 | Improved |
| use same proposal and inference settings for IoU-Net | IoU-Net · COCO 2017 test-dev | AP (%): 38.3 -> 38.2 | No clear effect |
| use simple Gaussian output model instead of direct regression | Faster-RCNN · COCO 2017 test-dev | AP (%): 37.2 -> 36.7 | Hurt |
| add more mixture components (2, 4, 8) | Gaussian mixture · COCO 2017 test-dev | AP (%): 37.1 -> 36.8 | No clear effect |
| replace ATOM IoU-Net module with energy-based regression | ATOM tracker · TrackingNet | Success (%): 70.3 -> 74.5 | Improved |
| replace ATOM IoU-Net module with energy-based regression | ATOM tracker · UAV123 | AUC (%): 65.0 -> 67.2 | Improved |
| add gradient-based refinement | Direct L2 regression · UTKFace age estimation | MAE: 4.81 -> 4.65 | Improved |
| add gradient-based refinement | Laplace · UTKFace age estimation | MAE: 4.85 -> 4.81 | No clear effect |
| add gradient-based refinement | Direct · BIWI head pose | average MAE: 3.09 -> 3.07 | No clear effect |
| use two-component Gaussian proposal instead of one (ablation) | proposal distribution · COCO 2017 val | AP (%): 38.47 -> 39.05 | Improved |
| use three components instead of two (ablation) | proposal distribution · COCO 2017 val | AP (%): 39.05 -> 39.11 | No clear effect |
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.
- ATOM — ATOM: Accurate Tracking by Overlap Maximization (CVPR 2019) · better on 7 of 7 shared results
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
- Probabilistic Regression for Visual Tracking (CVPR 2020 · similarity 0.89)
- Estimating and Evaluating Regression Predictive Uncertainty in Deep Object Detectors (ICLR 2021 · similarity 0.88)
- A Comprehensive Analysis of Deep Regression (TPAMI 2020 · similarity 0.88)
- Deep Probabilistic Supervision for Image Classification (NeurIPS 2026 · similarity 0.86)
- Learning normalized image densities via dual score matching (NeurIPS 2025 · similarity 0.86)
- Bayesian Exploration of Pre-trained Models for Low-shot Image Classification (CVPR 2024 · similarity 0.85)
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.
