Research map / Detection and segmentation / CNN features
A Unified Sequence Interface for Vision Tasks
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 |
|---|---|---|---|
| Increase input resolution from 640×640 to 1024×1024 (ablation) | Pix2Seq v2 multi-task model · COCO 2017 val object detection | AP: 44.2 -> 46.5 | Helped |
| Increase input resolution from 640×640 to 1024×1024 (ablation) | Pix2Seq v2 multi-task model · COCO 2017 val instance segmentation | AP: 36.9 -> 38.2 | Helped |
| Increase input resolution from 640×640 to 1024×1024 (ablation) | Pix2Seq v2 multi-task model · COCO 2017 val keypoint detection | AP: 65.0 -> 64.8 | Hurt |
| Increase input resolution from 640×640 to 1024×1024 (ablation) | Pix2Seq v2 multi-task model · COCO 2017 val captioning | BLEU: 34.3 -> 34.9 | Helped |
| Generate and average 8 mask samples instead of using a single sample (ablation) | Pix2Seq v2 instance segmentation · COCO instance segmentation | AP: ~6 AP better | Helped |
| Increase mask samples beyond 12 (ablation) | Pix2Seq v2 instance segmentation · COCO instance segmentation | AP | No clear effect |
| Evaluate segmentation using a crop containing the prompted object (ablation) | Pix2Seq v2 instance segmentation at 640×640 · COCO instance segmentation | AP: 1.3 AP improvement | Helped |
| Evaluate segmentation using a crop containing the prompted object (ablation) | Pix2Seq v2 instance segmentation at 1024×1024 · COCO instance segmentation | AP | No clear effect |
| Use a crop twice the person bounding-box size instead of an extremely large crop (ablation) | Pix2Seq v2 keypoint detection · COCO keypoint detection | AP: ~9 AP improvement | Helped |
| Use a crop twice the person bounding-box size instead of the exact bounding box (ablation) | Pix2Seq v2 keypoint detection · COCO keypoint detection | AP | Helped |
| Set invisible-keypoint token loss weight to 0.0 instead of 0.1 (ablation) | Pix2Seq v2 keypoint detection · COCO keypoint detection | AP: 12 AP lower | Hurt |
| Set invisible-keypoint token loss weight to 1.0 instead of 0.1 (ablation) | Pix2Seq v2 keypoint detection · COCO keypoint detection | AP: lower by 1 | Hurt |
Closest papers by meaning
- Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks (CVPR 2024 · similarity 0.88)
- InstructCV: Instruction-Tuned Text-to-Image Diffusion Models as Vision Generalists (ICLR 2024 · similarity 0.86)
- Explanatory Instructions: Towards Unified Vision Tasks Understanding and Zero-shot Generalization (ICML 2025 · similarity 0.86)
- Benchmarking Large Vision-Language Models on Fine-Grained Image Tasks: A Comprehensive Evaluation (ICLR 2026 · similarity 0.86)
- Masked AutoDecoder is Effective Multi-Task Vision Generalist (CVPR 2024 · similarity 0.86)
- Aligning and Prompting Everything All at Once for Universal Visual Perception (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.