Skip to content

Paper 076 / University of Washington / Allen Institute for AI

You Only Look Once: Unified, Real-Time Object Detection

YOLO reframes object detection as a single regression problem that predicts boxes and class probabilities directly from a full image.

UPDATED 16 Jul 2026SOURCE-LED REVIEWRESEARCH REVIEW
01

Paper, researchers, and primary source

Major lab research / computer_vision

YOLO reframes object detection as a single regression problem that predicts boxes and class probabilities directly from a full image.

CONTRIBUTION / 01

Contribution 1

Unified localization and classification in one end-to-end neural network.

CONTRIBUTION / 02

Contribution 2

Enabled real-time detection with a single image evaluation.

CONTRIBUTION / 03

Contribution 3

Analyzed characteristic localization and background-error tradeoffs against two-stage detectors.

02

Research context

computer_vision / 2015

You Only Look Once: Unified, Real-Time Object Detection places yolo inside the broader computer vision discussion at University of Washington / Allen Institute for AI, with object detection supplying a second analytical lens. Read together, the source records three advances: Unified localization and classification in one end-to-end neural network; Enabled real-time detection with a single image evaluation; and Analyzed characteristic localization and background-error tradeoffs against two-stage detectors. Keeping those moves together prevents real time vision from being detached from its evidence. For an implementation review, the relevant consequence is that A fixed grid and early detection design struggle with small or clustered objects, and results on benchmark images do not guarantee deployment robustness.

03

Methods and evidence reading

1 cataloged method notes

METHOD / 01

Method 1

The experimental design in You Only Look Once: Unified, Real-Time Object Detection tests unified localization and classification in one end-to-end neural network and enabled real-time detection with a single image evaluation against the paper's documented baselines, datasets, model variants, or systems workloads.

How to read the evidence

The evidentiary value of You Only Look Once: Unified, Real-Time Object Detection comes from the relationship among its reported moves. Two entries define the method-level claim: Unified localization and classification in one end-to-end neural network; and Enabled real-time detection with a single image evaluation. The cataloged result is: Analyzed characteristic localization and background-error tradeoffs against two-stage detectors. On that basis, A fixed grid and early detection design struggle with small or clustered objects, and results on benchmark images do not guarantee deployment robustness. The catalog nevertheless records that the claim attached to You Only Look Once: Unified, Real-Time Object Detection is conditional on selected metrics, prompt protocol, image resolution, reported datasets, input modalities, and domain coverage, so it cannot be generalized from the paper title alone. Reproduction work should separate genuine yolo transfer from behavior caused by a changed object detection setup. Rechecking You Only Look Once: Unified, Real-Time Object Detection calls for an explicit yolo baseline, controlled object detection changes, and a trace of cases that challenge Enabled real-time detection with a single image evaluation under the new setup.

04

Findings in the source record

1 paper-specific findings

  1. The reported evidence in You Only Look Once: Unified, Real-Time Object Detection supports analyzed characteristic localization and background-error tradeoffs against two-stage detectors.
05

Practical implication for AI builders

University of Washington / Allen Institute for AI / 2015

06

Proposed BrokenGPT application

Research blueprint / proposed status

INTEGRATION POINT / 01

Proposed route placement / yolo

For a proposed BrokenGPT experiment based on You Only Look Once: Unified, Real-Time Object Detection, use detection only behind a task-specific benchmark covering small objects, occlusion, lighting, false positives, and latency on deployment hardware. Keep the yolo path isolated, versioned, and attributable to this research record.

VALIDATION METRIC / 02

Proposed acceptance test / object detection

Validate the proposed yolo route against the paper's reported outcome: Analyzed characteristic localization and background-error tradeoffs against two-stage detectors. For the You Only Look Once: Unified, Real-Time Object Detection prototype, collect boundary errors, task accuracy, transfer robustness, and calibration and audit object detection slices independently before promoting the yolo configuration.

TRADEOFF / 03

Proposed decision boundary / real time vision

Balance resolution, latency, and domain shift before promoting the proposed real time vision design. Because A deployment review should isolate media rights, representation bias, memorization, source provenance, accessibility, and misuse when translating the yolo contribution into a different system, adoption remains conditional on replay under BrokenGPT's selected model, runtime, and policy configuration.

07

Limitations, verification, and source

Boundaries recorded with the paper

Limitations

  • The claim attached to You Only Look Once: Unified, Real-Time Object Detection is conditional on selected metrics, prompt protocol, image resolution, reported datasets, input modalities, and domain coverage, so it cannot be generalized from the paper title alone.
  • A deployment review should isolate media rights, representation bias, memorization, source provenance, accessibility, and misuse when translating the yolo contribution into a different system.

PRIMARY SOURCES

  1. 01
    You Only Look Once: Unified, Real-Time Object Detection

    University of Washington / Allen Institute for AI — Primary primary arXiv paper / 8 June 2015 / Joseph Redmon, Santosh Divvala, Ross Girshick, and 1 more

Related research reviews

View all 100 credited research papers

STRAIGHT ANSWERS

Frequently asked questions

01What does You Only Look Once: Unified, Real-Time Object Detection study?

YOLO reframes object detection as a single regression problem that predicts boxes and class probabilities directly from a full image.

02Which methods does You Only Look Once: Unified, Real-Time Object Detection use?

The experimental design in You Only Look Once: Unified, Real-Time Object Detection tests unified localization and classification in one end-to-end neural network and enabled real-time detection with a single image evaluation against the paper's documented baselines, datasets, model variants, or systems workloads.

03What does You Only Look Once: Unified, Real-Time Object Detection report?

The reported evidence in You Only Look Once: Unified, Real-Time Object Detection supports analyzed characteristic localization and background-error tradeoffs against two-stage detectors.

04What is the proposed BrokenGPT application for You Only Look Once: Unified, Real-Time Object Detection?

Proposed: use detection only behind a task-specific benchmark covering small objects, occlusion, lighting, false positives, and latency on deployment hardware.

MAJOR LAB RESEARCH / PAPER 076

Continue after You Only Look Once: Unified, Real-Time Object Detection

After You Only Look Once: Unified, Real-Time Object Detection, browse the full index for adjacent computer_vision research and work from University of Washington / Allen Institute for AI.

Open the paper index