FFFrancesco Di Felice

Francesco Di Felice

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Francesco Di Felice has a total of 9 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • InstantPose: Zero-Shot Instance-Level 6D Pose Estimation From a Single View

    by Francesco Di Felice, Alberto Remus, Stefano Gasperini, Benjamin Busam, Lionel Ott, Stefan Thalhammer, Federico Tombari, Carlo Alberto Avizzano

    Object pose estimation using visual data is crucial for robotic interaction with the environment. Many existing instance-level methods are restricted by their requirements for 3D CAD models or multiple object views, which limits their flexibility and generalizability. Overcoming this limitation is critical to enhance the adaptability of pose estimation systems. In this work, a novel pipeline that leverages recent advances in reconstruction techniques is presented to address these challenges. To this end, Large Reconstruction Models (LRM) represent an advanced neural architecture capable of generating 3D object models from a limited set of views. Nevertheless, the resulting 3D models often lack relevant geometric and texture details due to insufficient input information. This research presents InstantPose, an innovative zero-shot instance-level pose estimation method that, building upon LRM, can determine the pose of unseen objects using as little as a single unposed RGB reference and RGB-D query images. Extensive experiments demonstrate that InstantPose achieves remarkable performance in object pose estimation on the YCB-V dataset, compared to methods conceived to rely on a geometrically perfect object's model. Furthermore, the 6D pose provided through the presented approach facilitates successful object grasping, highlighting its practical utility in robotic manipulation tasks.


  • Open Access

    One-Shot Imitation Learning With Graph Neural Networks for Pick-and-Place Manipulation Tasks

    by Francesco Di Felice, Salvatore D'AVella, Alberto Remus, Paolo Tripicchio, Carlo Alberto Avizzano

    The proposed work presents a framework based on Graph Neural Networks (GNN) that abstracts the task to be executed and directly allows the robot to learn task-specific rules from synthetic demonstrations given through imitation learning. A graph representation of the state space is considered to encode the task-relevant entities as nodes for a Pick-and-Place task declined at different levels of difficulty. During training, the GNN-based policy learns the underlying rules of the manipulation task focusing on the structural relevance and the type of objects and goals, relying on an external primitive to move the robot to accomplish the task. The GNN-policy has been trained as a node-classification approach by looking at the different configurations of the objects and goals present in the scene, learning the association between them with respect to their type for the Pick-and-Place task. The experimental results show a high generalization capability of the proposed model in terms of the number, positions, height distributions, and even configurations of the objects/goals. Thanks to the generalization, only a single image of the desired goal configuration is required at inference time.


  • Open Access

    i2c-net: Using Instance-Level Neural Networks for Monocular Category-Level 6D Pose Estimation

    by Alberto Remus, Salvatore D'Avella, Francesco Di Felice, Paolo Tripicchio, Carlo Alberto Avizzano

    Object detection and pose estimation are strict requirements for many robotic grasping and manipulation applications to endow robots with the ability to grasp objects with different properties in cluttered scenes and with various lighting conditions. This work proposes the framework i2c-net to extract the 6D pose of multiple objects belonging to different categories, starting from an instance-level pose estimation network and relying only on RGB images. The network is trained on a custom-made synthetic photo-realistic dataset, generated from some base CAD models, opportunely deformed, and enriched with real textures for domain randomization purposes. At inference time, the instance-level network is employed in combination with a 3D mesh reconstruction module, achieving category-level capabilities. Depth information is used for post-processing as a correction. Tests conducted on real objects of the YCB-V and NOCS-REAL datasets outline the high accuracy of the proposed approach.