Research map / Image generation / Autoregressive and latent generation
Pre-Trained Image Processing Transformer
Check 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 |
|---|---|---|---|
| add contrastive learning with λ=0.1 (ablation) | IPT · Set4, ×2 super-resolution | PSNR: 38.27 -> 38.37 | Helped |
| set contrastive-loss weight to λ=0.05 rather than λ=0 (ablation) | IPT · Set4, ×2 super-resolution | PSNR: 38.27 -> 38.32 | Helped |
| set contrastive-loss weight to λ=0.2 rather than λ=0 (ablation) | IPT · Set4, ×2 super-resolution | PSNR: 38.27 -> 38.33 | Helped |
| set contrastive-loss weight to λ=0.5 rather than λ=0 (ablation) | IPT · Set4, ×2 super-resolution | PSNR: 38.27 -> 38.26 | Hurt |
| use learnable position embeddings instead of fixed or no position embeddings (ablation) | IPT · Image processing tasks | PSNR: 0.2dB to 0.3dB | Helped |
| add task embeddings (ablation) | IPT · Image processing tasks | accuracy: 0.1dB to 0.5dB | Helped |
| use multi-task pre-training rather than single-task pre-training (ablation) | IPT · Pre-trained tasks and transfer to other tasks | PSNR: about 0.3dB | Mixed |
| increase synthesized ImageNet pre-training data to a large-scale fraction (ablation) | Transformer-based IPT versus CNN models · Image processing tasks | performance | Helped |
Closest papers by meaning
- An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale (ICLR 2021 · similarity 0.86)
- ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias (NeurIPS 2021 · similarity 0.85)
- Boosting Image Quality Assessment through Efficient Transformer Adaptation with Local Feature Enhancement (CVPR 2024 · similarity 0.84)
- Your ViT is Secretly an Image Segmentation Model (CVPR 2025 · similarity 0.84)
- DeiT-LT: Distillation Strikes Back for Vision Transformer Training on Long-Tailed Datasets (CVPR 2024 · similarity 0.83)
- Efficient Adaptation of Pre-trained Vision Transformer via Householder Transformation (NeurIPS 2024 · similarity 0.83)
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.