A Minimal Optical-Flow Representation for Vision-Based Tactile Rotation Classification
Overview
A study explores using compact optical-flow data to classify object rotation in robotic manipulation across different gravity conditions. Earth-trained models show 75.9% accuracy in orbital conditions, while a multi-domain model achieves 96.3% accuracy, emphasizing the need for domain adaptation in training data.
Generated from attributed reports · Updated 2 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · feature count: 126 other · Basis not reported
“ted over a 7x9 grid into 126 features and used to classify the direction of load-induced rotation under Earth, Mars, Moon, and orbital gravity. Gravity causes a small but significant shift in these features, accounting for 1.6% of their variance (R2 = 0.016). Despite its small magnitude, this shift affects models trained only on Earth data: XGBoost accuracy decreases from 94.4% on Earth to 75.9% in orbit. In contrast”
Exact source · revision 1Source owner not reported
Reported quantity · Reported accuracy: 94.4 percent · Basis not reported · Differing source assertions
“ted over a 7x9 grid into 126 features and used to classify the direction of load-induced rotation under Earth, Mars, Moon, and orbital gravity. Gravity causes a small but significant shift in these features, accounting for 1.6% of their variance (R2 = 0.016). Despite its small magnitude, this shift affects models trained only on Earth data: XGBoost accuracy decreases from 94.4% on Earth to 75.9% in orbit. In contrast”
Exact source · revision 1Source owner not reported
Reported quantity · Reported accuracy: 75.9 percent · Basis not reported · Differing source assertions
“ted over a 7x9 grid into 126 features and used to classify the direction of load-induced rotation under Earth, Mars, Moon, and orbital gravity. Gravity causes a small but significant shift in these features, accounting for 1.6% of their variance (R2 = 0.016). Despite its small magnitude, this shift affects models trained only on Earth data: XGBoost accuracy decreases from 94.4% on Earth to 75.9% in orbit. In contrast”
Exact source · revision 1Source owner not reported
Reported quantity · Reported accuracy: 95.7 percent · Basis not reported · Differing source assertions
“, a single model trained across all four gravity domains achieves 96.3% overall accuracy and 95.1%-97.0% across individual domains, without using gravity as an input. The representation can also be reduced to 40 features while retaining 95.7% accuracy, with XGBoost requiring only 0.14 ms per inference. These findings show that Earth-gravity performance alone is insufficient to establish the transferability of tactile”
Exact source · revision 1Source owner not reported
Reported quantity · inference time for XGBoost: 0.14 other · Basis not reported
“, a single model trained across all four gravity domains achieves 96.3% overall accuracy and 95.1%-97.0% across individual domains, without using gravity as an input. The representation can also be reduced to 40 features while retaining 95.7% accuracy, with XGBoost requiring only 0.14 ms per inference. These findings show that Earth-gravity performance alone is insufficient to establish the transferability of tactile”
Exact source · revision 1Source owner not reported
Reported quantity · Reported accuracy: 95.1 percent · Basis not reported · Differing source assertions
“, a single model trained across all four gravity domains achieves 96.3% overall accuracy and 95.1%-97.0% across individual domains, without using gravity as an input. The representation can also be reduced to 40 features while retaining 95.7% accuracy, with XGBoost requiring only 0.14 ms per inference. These findings show that Earth-gravity performance alone is insufficient to establish the transferability of tactile”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsA Minimal Optical-Flow Representation for Vision-Based Tactile Rotation Classification in Robotic Manipulation Across Gravity Domains
This paper investigates the use of a compact optical-flow representation for classifying object rotation in robotic manipulation across different gravity conditions. Using simulated GelSight Mini data, the study shows that gravity-induced shifts in tactile features affect model performance, with Earth-trained models achieving only 75.9% accuracy in orbital conditions. A model trained across all four gravity domains achieves 96.3% overall accuracy, highlighting the need to account for domain shifts during training.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).