Research map / Image generation / Autoregressive and latent generation
Multistage Adversarial Losses for Pose-Based Human Image Synthesis
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the pose-based multistage method instead of PG2 | Human image synthesis · Human3.6M test | SSIM: 0.60 -> 0.72 | Helped |
| Use the pose-based multistage method instead of PG2 | Human image synthesis · Human3.6M test | PSNR: 19.19 -> 20.62 | Helped |
| Add an image-only discriminator to Unet (ablation) | Foreground transformer · Human3.6M test foreground | SSIM: 0.77 -> 0.73 | Hurt |
| Add an image-only discriminator to Unet (ablation) | Foreground transformer · Human3.6M test foreground | PSNR: 20.49 -> 19.39 | Hurt |
| Add a pose encoder to Unet without a discriminator (ablation) | Foreground transformer · Human3.6M test foreground | SSIM: 0.77 -> 0.82 | Helped |
| Add a pose encoder to Unet without a discriminator (ablation) | Foreground transformer · Human3.6M test foreground | PSNR: 20.49 -> 22.57 | Helped |
| Add skip connections to the pose- and GAN-equipped encoder-decoder (ablation) | Foreground transformer · Human3.6M test foreground | SSIM: 0.75 -> 0.81 | Helped |
| Add skip connections to the pose- and GAN-equipped encoder-decoder (ablation) | Foreground transformer · Human3.6M test foreground | PSNR: 19.81 -> 22.10 | Helped |
| Add a GAN discriminator to Unet+Pose (ablation) | Foreground transformer · Human3.6M test foreground | SSIM: 0.82 -> 0.81 | Hurt |
| Add a GAN discriminator to Unet+Pose (ablation) | Foreground transformer · Human3.6M test foreground | PSNR: 22.57 -> 22.10 | Hurt |
| Replace the background discriminator with a global discriminator while retaining the foreground discriminator (ablation) | Background transformer · Human3.6M test background | SSIM: 0.67 -> 0.72 | Helped |
| Replace the background discriminator with a global discriminator while retaining the foreground discriminator (ablation) | Background transformer · Human3.6M test background | PSNR: 19.70 -> 20.62 | Helped |
Closest papers by meaning
- Unsupervised Person Image Synthesis in Arbitrary Poses (CVPR 2018 · similarity 0.89)
- Deformable GANs for Pose-Based Human Image Generation (CVPR 2018 · similarity 0.87)
- Person Image Synthesis via Denoising Diffusion Model (CVPR 2023 · similarity 0.87)
- Pose Guided Person Image Generation (NeurIPS 2017 · similarity 0.87)
- A Dual-Augmentor Framework for Domain Generalization in 3D Human Pose Estimation (CVPR 2024 · similarity 0.85)
- Towards Robust 3D Pose Transfer with Adversarial Learning (CVPR 2024 · 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.