AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

X-World: Accessibility, Vision, and Autonomy Meet

Jimuyang Zhang, Minglan Zheng, Matthew Boyd, Eshed Ohn-Bar

ICCV 2021 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Vision-language navigation agents · 18 citations (OpenAlex)

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 changedWhereMeasuredEffect
fine-tune on X-World simulated dataCOCO-pretrained Mask R-CNN · X-World Sim test (person/vehicle model)AP: 35.1 -> 59.6Improved
fine-tune on X-World simulated dataCityscapes-pretrained Mask R-CNN · X-World Sim test (person/vehicle model)AP: 25.1 -> 60.0Improved
fine-tune on X-World simulated dataCOCO-pretrained Mask R-CNN · X-World Sim test (person/vehicle model), person-wc classAP: 8.9 -> 31.3Improved
COCO pretraining instead of CityscapesMask R-CNN · X-World Sim test (person/vehicle model)AP: 25.1 -> 35.1Improved
COCO pretraining instead of Cityscapes (after fine-tuning)Mask R-CNN · X-World Sim test (person/vehicle model)AP: 59.6 vs 60.0No clear effect
new mobility aids test setting added to test (ablation)COCO & X-World (Sim) fine-grained · X-World Sim fine-grained 8 classesAP: 43.9 -> 39.9Hurt
new weathers test setting addedCOCO & X-World (Sim) fine-grained · X-World Sim fine-grained 8 classesAP: 43.9 -> 43.7No clear effect
train on real data only vs simulation onlyCOCO & X-World · X-World Real-world benchmarkAP: 6.6 -> 29.8Improved
add simulated data to real training dataCOCO & X-World (Real) · X-World Real-world benchmark, wc classAP: 33.7 -> 36.1Improved
add simulated data to real training dataCOCO & X-World (Real) · X-World Real-world benchmark, cane classAP: 0.66 -> 1.83Improved
add simulated data to real training dataCOCO & X-World (Real) · X-World Real-world benchmark, person-wc classAP: 28.2 -> 30.3Improved
add simulated data to real training dataCityscapes & X-World (Real) · X-World Real-world benchmark, cane classAP: 0.70 -> 1.41Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy