Welcome to ISPGroup!

Accepted paper: “MPL: Lifting 3D Human Pose from Multi-view 2D Poses” by Seyed Abolfazl Ghasemzadeh, A. Alahi, and C. De Vleeschouwer, to be presented at “T-CAP: Towards a Complete Analysis of People: Fine-grained Understanding for Real-World Applications” workshop at the European Conference on Computer Vision (ECCV, Milan, 1-4 Oct 2024).

MPL is a 3D human pose lifter that takes multi-view 2D poses as its inputs. The main challenge in multi-view 3D human pose estimation is lack of real-life images paired with 3D. This challenge is bypassed in MPL by relying on a two-stage framework consisting of an off-the-shelf 2D pose estimator and a multi-view 3D pose lifter. More information here

Accepted paper: “Sequential Representation Learning via Static-Dynamic Conditional Disentanglement " by Mathieu Cyrille Simon, Pascal Frossard and Christophe De Vleeschouwer (arXiv preprint), to be presented at the European Conference on Computer Vision (ECCV, Milan, 1-4 Oct 2024).

Unsupervised disentangled representation learning within sequential data, focusing on separating time-independent and time-varying factors in videos. This disentanglement is expected to facilitate numerous downstream generation or classification tasks and enhance model explainability.

Accepted paper: “Keypoint Promptable Re-Identification” by Vladimir Somers, A. Alahi and C. De Vleeschouwer (arXiv preprint), to be presented at the European Conference on Computer Vision (ECCV, Milan, 1-4 Oct 2024).

KPR is SOTA re-identification method, with optional keypoint prompts, that is robust to multi-person occlusions and plug-and-play with any pose estimator (or through manual keypoint selection). More information here.

Best Poster Award at the International Conference on Space Robotics (ISpaRo24): Antoine Legrand, Renaud Detry and Christophe De Vleeschouwer, “Leveraging neural radiance fields for pose estimation of an unknown space object during proximity operations” (arXiv), ISpaRo 2024.