AnyViewDex: View-Invariant Dexterous Manipulation from RGB Observations
为解决多指灵巧操作对相机视角变化敏感的问题，本文提出AnyViewDex方法，通过在模拟训练中编码几何知识来实现仅依赖单目RGB图像的视点不变控制。
0 citationsRead paper
Selected work
Representative Papers
Modeling What Changes: Sparse, Residual World Models for Object-Centric Manipulation
本文提出了一种稀疏/残差模型，通过仅对场景中变化的对象进行建模来提高物理预测和控制的准确性与可解释性，解决了传统整体世界模型重复预测静态场景部分的问题。
0 citationsRead paper
Devising a solution to the problems of Cancer awareness in Telangana
In Telangana, India, extremely low early-screening rates for cervical (3.3%), breast (0.3%), and oral (2.3%) cancers—coupled with poor public awareness—contribute to high incidence and mortality. Method: This study develops lightweight, demography-based machine learning models—decision trees for cervical cancer and SVMs for breast cancer—and integrates them with geofence-driven local healthcare facility recommendation (LBS), dynamic electronic health card management, and community-targeted health education via mobile/web platforms. Contribution/Results: It is the first system in the region to unify risk prediction, intelligent referral, longitudinal health record creation, and precision health communication into a closed-loop framework. The platform covers over 1,200 certified cancer centers statewide; predictive models achieve AUCs of 0.82–0.87. Empirical evaluation demonstrates >40% improvement in cancer symptom recognition among target populations.
0 citationsRead paper
Recent publications
Latest Papers
AnyViewDex: View-Invariant Dexterous Manipulation from RGB Observations
为解决多指灵巧操作对相机视角变化敏感的问题，本文提出AnyViewDex方法，通过在模拟训练中编码几何知识来实现仅依赖单目RGB图像的视点不变控制。
0 citationsRead paper
Modeling What Changes: Sparse, Residual World Models for Object-Centric Manipulation
本文提出了一种稀疏/残差模型，通过仅对场景中变化的对象进行建模来提高物理预测和控制的准确性与可解释性，解决了传统整体世界模型重复预测静态场景部分的问题。
0 citationsRead paper
Devising a solution to the problems of Cancer awareness in Telangana
In Telangana, India, extremely low early-screening rates for cervical (3.3%), breast (0.3%), and oral (2.3%) cancers—coupled with poor public awareness—contribute to high incidence and mortality. Method: This study develops lightweight, demography-based machine learning models—decision trees for cervical cancer and SVMs for breast cancer—and integrates them with geofence-driven local healthcare facility recommendation (LBS), dynamic electronic health card management, and community-targeted health education via mobile/web platforms. Contribution/Results: It is the first system in the region to unify risk prediction, intelligent referral, longitudinal health record creation, and precision health communication into a closed-loop framework. The platform covers over 1,200 certified cancer centers statewide; predictive models achieve AUCs of 0.82–0.87. Empirical evaluation demonstrates >40% improvement in cancer symptom recognition among target populations.
0 citationsRead paper