Research map / Detection and segmentation / Open-set and few-shot detection
VTNet: Visual Transformer Network for Object Goal Navigation
Heming Du, Xin Yu, Liang Zheng
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 |
|---|---|---|---|
| use VTNet instead of Baseline | AI2-Thor navigation · AI2-Thor test, ALL | Success: 62.6 -> 72.2 | Improved |
| use VTNet instead of ORG+TPN | AI2-Thor navigation · AI2-Thor test, ALL | Success: 69.3 -> 72.2 | Improved |
| add global features (ablation) | VTNet · AI2-Thor test, ALL | Success: 67.0 -> 72.2 | Improved |
| add the VT decoder to attend global descriptors to local descriptors (ablation) | VTNet · AI2-Thor test, ALL | Success: 67.0 -> 72.2 | Improved |
| add positional embedding to global features (ablation) | VTNet · AI2-Thor test, ALL | Success: 71.0 -> 72.2 | Improved |
| use VTNet's decoder arrangement instead of feeding positional global features into the encoder (ablation) | VTNet · AI2-Thor test, ALL | Success: 70.1 -> 72.2 | Improved |
| use DETR instance features instead of Faster R-CNN features (ablation) | VTNet · AI2-Thor test, ALL | Success: 70.1 -> 72.2 | Improved |
| use DETR instance features instead of Faster R-CNN† features (ablation) | VTNet · AI2-Thor test, ALL | Success: 70.3 -> 72.2 | Improved |
| pre-train VT with optimal action instructions (ablation) | VTNet · AI2-Thor navigation | navigation policy effectiveness | Improved |
| use 2 encoder and 2 decoder layers instead of 1 each (ablation) | VTNet · AI2-Thor test | Success: 71.2 -> 72.2 | Improved |
| use 2 encoder and 2 decoder layers instead of 4 each (ablation) | VTNet · AI2-Thor test | Success: 70.0 -> 72.2 | Improved |
| use 4 attention heads instead of 8, with 2 encoder and decoder layers (ablation) | VTNet · AI2-Thor test | Success: 70.9 -> 72.2 | Improved |
Later papers that compared with this method
1 later accepted paper reports VTNet as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Multiple Thinking Achieving Meta-Ability Decoupling for Object Navigation (ICML 2023 · better on 1 of 1 shared result)
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
- Learning Object Relation Graph and Tentative Policy for Visual Navigation (ECCV 2020 · similarity 0.89)
- Visual Graph Memory With Unsupervised Representation for Visual Navigation (ICCV 2021 · similarity 0.88)
- Generative Meta-Adversarial Network for Unseen Object Navigation (ECCV 2022 · similarity 0.87)
- Towards Versatile Embodied Navigation (NeurIPS 2022 · similarity 0.87)
- End-to-End (Instance)-Image Goal Navigation through Correspondence as an Emergent Phenomenon (ICLR 2024 · similarity 0.87)
- Seeing through Uncertainty: Robust Task-Oriented Optimization in Visual Navigation (NeurIPS 2025 · 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.
