Research map / Detection and segmentation / Open-set and few-shot detection
Pix2seq: A Language Modeling Framework for Object Detection
Ting Chen, Saurabh Saxena, Lala Li, David J. Fleet, Geoffrey Hinton
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 |
|---|---|---|---|
| Pretrain Pix2Seq on Objects365 before COCO fine-tuning | Pix2Seq · COCO 2017 val | AP: 45.0 -> 50.0 | Improved |
| Increase fine-tuning image size from 640×640 to 1333×1333 | Objects365-pretrained Pix2Seq, R50 · COCO 2017 val | AP: 39.1 -> 42.6 | Improved |
| Add sequence augmentation during fine-tuning (ablation) | Pix2Seq pretrained on Objects365 · COCO 2017 val | AR@100: 55.4 -> 61.6 | Improved |
| Add sequence augmentation during fine-tuning (ablation) | Pix2Seq pretrained on Objects365 · COCO 2017 val | AP: 43.7 -> 44.5 | Improved |
| Add sequence augmentation during pretraining as well as fine-tuning (ablation) | Pix2Seq pretrained on Objects365 · COCO 2017 val | AR@100: 61.6 -> 61.7 | No clear effect |
| Use deterministic object ordering instead of random ordering (ablation) | Pix2Seq sequence construction · COCO | AP | Hurt |
| Use fewer than 500 coordinate quantization bins (ablation) | Pix2Seq sequence construction · COCO, images with longest side 640 pixels | AP | Hurt |
| Use nucleus sampling instead of argmax sampling at inference (ablation) | Pix2Seq · COCO | AP | Improved |
| Remove sequence augmentation and delay EOS sampling (ablation) | Pix2Seq trained from scratch on COCO · COCO | AR | Hurt |
Later papers that compared with this method
2 later accepted papers report Pix2Seq as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- DiffuDETR: Rethinking Detection Transformers with Diffusion Process (ICLR 2026 · better on 1 of 1 shared result)
- Obj2Seq: Formatting Objects as Sequences with Class Prompt for Visual Tasks (NeurIPS 2022 · better on 2 of 2 shared results)
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.
- Faster R-CNN+ — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 2 of 2 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
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- Generative Region-Language Pretraining for Open-Ended Object Detection (CVPR 2024 · similarity 0.88)
- InstructDET: Diversifying Referring Object Detection with Generalized Instructions (ICLR 2024 · similarity 0.88)
- ROD-MLLM: Towards More Reliable Object Detection in Multimodal Large Language Models (CVPR 2025 · similarity 0.88)
- DoGA: Enhancing Grounded Object Detection via Grouped Pre-Training with Attributes (AAAI 2025 · similarity 0.88)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.87)
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.
