--- name: ieee-transactions-on-pattern-analysis-and-machine-intelligence description: Use when targeting IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) or deciding whether a computer vision or machine learning manuscript fits this archival IEEE venue. Encodes the journal's fit, framing, method-and-evidence bar, house style, official-submission re-check, and desk-reject heuristics. --- # IEEE Transactions on Pattern Analysis and Machine Intelligence (ieee-transactions-on-pattern-analysis-and-machine-intelligence) ## Journal positioning IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) is the IEEE Computer Society's flagship archival journal for computer vision, pattern recognition, and machine learning methods. It is a journal — not a conference proceeding — and its culture reflects that distinction: papers are expected to be definitive, complete, and archival in depth rather than fast-moving and preliminary. The readership is the technical computer vision and ML methods community, and the bar is methodological rigor, thorough evaluation, and reproducibility rather than the hype-and-benchmark culture of conference proceedings. Accepted papers often represent a mature, extended contribution relative to prior conference work. This skill is a **fit / venue-selection / re-framing** tool. It does not replace the journal's current official submission guidelines. Before submitting, re-check the live author instructions on the IEEE Computer Society / TPAMI site and the submission system. ## When to trigger - The author names TPAMI as the target venue. - A computer vision, pattern recognition, or ML methods paper is ready for archival publication and the author is assessing depth, reproducibility, and format fit. - A conference paper is being extended to journal length and the author wants to ensure the extension meets TPAMI's archival standard. - The author needs to understand TPAMI's specific rejection profile compared with conference venues. ## Scope & topic fit - Computer vision: image/video understanding, 3-D reconstruction, segmentation, object detection, recognition, and generation — with methodological advance. - Pattern recognition: statistical and structural pattern analysis methods with demonstrated generality across problems or domains. - Machine learning methods with direct relevance to vision or broader ML foundations: representation learning, deep architectures, optimization, generalization theory. - Medical image analysis, remote sensing, document analysis, and scene understanding where methodological rigor is primary. - Evaluation methodology, datasets, and benchmarking — where the contribution is the benchmark design and analysis, not just performance on it. - Multimodal learning, video understanding, and 3-D vision methods. ## Method & evidence bar - Archival completeness is the core standard: the paper must be self-contained, with thorough derivations, proofs of key properties (where applicable), and complete experimental protocols. - Evaluation must be exhaustive by journal standards: multiple standard benchmarks, ablation studies that isolate each component's contribution, comparison against a broad set of contemporaneous baselines. - Reproducibility is not optional: code release is strongly expected; all hyperparameters, training details, and experimental conditions must be reported to enable reproduction. - Conference-extension papers must offer at least substantial additional technical content (new experiments, extended theory, new application domains, or deeper analysis) beyond the conference version; re-check current TPAMI extension policy. - Statistical rigor in comparison: performance differences should be assessed with appropriate significance tests where dataset size permits; cherry-picked examples should be accompanied by quantitative results. - Theoretical contributions (if any) must be rigorously proved; claims of theoretical properties require formal statements and proofs. ## Structure & house style - IEEE double-column format; re-check current TPAMI formatting requirements and length limits. - Abstract is unstructured; index terms (IEEE taxonomy) are required. - Introduction must clearly state the problem, existing limitations, the paper's technical contribution, and the extent of the evaluation. - A prior-work section that is thorough, fair, and analytically positioned — not a laundry list — is expected given the archival nature. - Experimental section should be organized by benchmark or research question, not by ablation type; make it easy to find the comparison a reviewer will want. - Supplementary material can carry additional experiments; main text should contain all central results. - For conference extensions, a section explicitly describing what is new relative to the conference version is required. ## Official-submission checklist - Before giving submission-ready advice, read `../../resources/source-basis.md` and `../../resources/official-source-map.md`; start from the official source anchors for this journal family, then cite the current journal-specific page you checked. - Search the live IEEE site for "TPAMI information for authors" and follow the current IEEE Computer Society version. - Re-check page/length limits and double-column formatting requirements. - Confirm IEEE Open Access options and check the current policy on conference-to-journal extensions. - Prepare index terms using IEEE taxonomy. - Include code availability statement; check IEEE data-sharing and software-licensing requirements. - Complete IEEE conflict-of-interest and author-contribution disclosures. - Check AI-use disclosure requirements. - If the live official instructions conflict with this skill, the official instructions win. ## Pre-submission self-check - [ ] One sentence stating the methodological contribution and what CV/ML problem it definitively solves or advances. - [ ] The paper is archivally complete: proofs, derivations, and full experimental details are present or in supplementary material. - [ ] All baselines are contemporaneous and fairly compared; ablations isolate each design decision. - [ ] Code is released or committed to release; all training and evaluation details are reproducible. - [ ] If a conference extension, the new content is substantial and explicitly documented. - [ ] IEEE formatting, index terms, and length comply with current instructions. ## Common desk-reject triggers - Conference paper submitted without substantial new content; re-use of conference text without adequate extension. - Evaluation on a narrow set of benchmarks without ablations or without comparison to the full relevant prior art. - No code release and no compelling justification; results that cannot be reproduced from the paper alone. - Theoretical claims made without proof or with only intuitive justification. - Paper reads as a proceedings contribution (fast-paced, preliminary) rather than an archival journal article (complete, definitive). ## Re-routing decision Papers with broad ML significance beyond CV/pattern-recognition methods → `journal-of-machine-learning-research` or `nature-machine-intelligence`. Work with a strong robotics integration and demonstrated system capability → `science-robotics`. Fast-moving results where time-to-publication is critical → conference venues (CVPR, ICCV, NeurIPS). Pure ML theory without CV relevance → `journal-of-machine-learning-research`. ## Output format ```text [Fit] High / Medium / Low (one-line reason) [Target] IEEE Transactions on Pattern Analysis and Machine Intelligence [Topic tags] <2–3 closest topics> [Method/evidence] [Top risk] [Official items to re-check] [Re-route suggestion] ```