Research map / Image generation / Editing and style transfer
Masked Linear Regression for Learning Local Receptive Fields for Facial Expression Synthesis
Nazar Khan, Arbish Akram, Arif Mahmood, Sania Ashraf, Kashif Murtaza
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 masked local receptive fields instead of dense l2 regression | ridge regression · CK+ expression synthesis 12 mappings (grayscale) | mean MSE x100: 1.904 -> 1.460 | Improved |
| replace l1 sparsity with masked regression | masked regression · CK+ expression synthesis 12 mappings (grayscale) | mean MSE x100: 1.548 -> 1.460 | Improved |
| replace l0 sparsity with masked regression | masked regression · CK+ expression synthesis 12 mappings (grayscale) | mean MSE x100: 1.665 -> 1.460 | Improved |
| use 5x5 receptive field instead of 3x3 (ablation) | masked regression · CK+ expression synthesis 12 mappings (grayscale) | mean MSE x100: 1.487 -> 1.460 | Improved |
| use 9x9 receptive field instead of 5x5 (ablation) | masked regression · CK+ expression synthesis 12 mappings (grayscale) | mean MSE x100: 1.460 -> 1.548 | Hurt |
| use 7x7 receptive field instead of 5x5 (ablation) | masked regression · CK+ expression synthesis 12 mappings (grayscale) | mean MSE x100: 1.460 -> 1.501 | Hurt |
| apply masked regression per channel to RGB | masked regression · Expression synthesis RGB 56x56 12 mappings | mean MSE x100: 0.907 -> 0.862 | Improved |
| use masked regression instead of l1 for training speed | masked regression · RGB expression mappings | training time (s): 16.782 -> 0.010 | Improved |
| use masked regression instead of GANs for model size | masked regression · 128x128 images | model size (x10^4 parameters): 4100 -> 16.2 | Improved |
| use masked regression instead of Pix2Pix for speed | masked regression · 128x128 images | execution time (msec): 320 -> 2.70 | Improved |
| use masked regression instead of Pix2Pix on out-of-dataset images | masked regression · out-of-dataset images | accuracy drop (pct points): 35.72 -> 12.39 | Improved |
| learned bias added to weight transformation (ablation) | masked regression · qualitative synthesis | visual quality | Improved |
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.
- StarGAN — StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation (CVPR 2018) · better on 3 of 3 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
- Joint Pose and Expression Modeling for Facial Expression Recognition (CVPR 2018 · similarity 0.89)
- Controllable Image-to-Video Translation: A Case Study on Facial Expression Generation (AAAI 2019 · similarity 0.87)
- SynFER: Towards Boosting Facial Expression Recognition with Synthetic Data (ICCV 2025 · similarity 0.87)
- 3D Facial Expressions through Analysis-by-Neural-Synthesis (CVPR 2024 · similarity 0.87)
- TEASER: Token Enhanced Spatial Modeling for Expressions Reconstruction (ICLR 2025 · similarity 0.86)
- Learning Adaptive Spatial Coherent Correlations for Speech-Preserving Facial Expression Manipulation (CVPR 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.
