AI Paper Atlas

Research map / Detection and segmentation / Weakly supervised detection and segmentation

Look Around and Learn: Self-Training Object Detection by Exploration

Gianluca Scarpellini, Stefano Rosa, Pietro Morerio, Lorenzo Natale, Alessio Del Bue

ECCV 2024 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Weakly supervised detection · 8 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
use Look Around policy with Disagreement Reconciliation instead of off-the-shelf detectorMask R-CNN · Gibson simulatedmAP@50: 40.33 -> 46.60Improved
use Disagreement Reconciliation perception instead of self-training with random goalsRandom policy · Gibson simulatedmAP@50: 39.67 -> 41.88Improved
use plain self-training with the Look Around policyLook Around · Gibson simulatedmAP@50: 40.33 -> 38.66Hurt
add instance-matching triplet loss (ablation)Look Around + Disag. reconc. · Gibson simulatedmAP@50: 44.19 -> 46.60Improved
add soft distillation loss with alpha 0.7 (ablation)Look Around + Disag. reconc. · Gibson simulatedmAP@50: 41.09 -> 46.60Improved
set distillation weight alpha to 1.0 instead of 0.7 (ablation)Look Around + Disag. reconc. · Gibson simulatedmAP@50: 46.60 -> 41.45Hurt
add soft distillation loss with Semantic Curiosity policy (ablation)Semantic Curiosity + Disag. reconc. · Gibson simulatedmAP@50: 43.16 -> 41.37Hurt
use entropy-based disagreement score instead of cosine, Euclidean or class-count scores (ablation)Look Around policy reward · Gibson simulatedmAP@50: 45.65 / 42.44 / 40.58 -> 46.60Improved
use Look Around instead of SEAL pipelineexploration policy · Gibson simulatedmAP@50: +3.59% over SEAL, +2.45% over Informative TrajectoriesImproved
use SEAL perception instead of Disagreement Reconciliation with SEAL policySEAL policy · Gibson simulatedmAP@50: 43.01 vs 42.38; Inform 44.15 vs 43.70Mixed
deploy Look Around on a real robot with Disagreement ReconciliationMask R-CNN · Real robot testmAP@50: 55.83 -> 65.62Improved
use human-guided goals instead of learned policyreal robot data collection · Real robot testmAP@50: 65.62 -> 56.00Hurt

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