Research

​​​​​​

Multimodal scene understanding & Analysis and understanding of visual content 

• Semantic image inpainting through improved Wasserstein generative adversarial networks
• Towards inverse problems with multiple solutions: An Analysis of generative methods for multiple image inpainting
• The Art of Inpainting (book)

• Anomaly detection 
  • Slides of a related talk

• Adversarial picture colorization  +  code   +   demo
• DIP-VBTV: a color image restoration model  + code
• Analysis of different losses for deep learning image colorization + of color spaces

• Joint multi-view blind restoration
• Real-Time High-Resolution Image Inpainting on Edge Devices
• Photorealistic facial wrinkle removal

• Multimodal scene understanding
• Self-supervised video summarization :  TRIM  +  TRIMMER using reinforcement learning

• Sign language translation and generation

• Visual motif identification: Elaboration of a curated comparative dataset and classification methods

 ​​​

​​​

Multimodal scene understanding & Geometric and semantic sports analysis (slides of a related talk)

• Always Look on the Bright Side of the Field  +  Deep learning model for estimating orientation directly from video footage
• Using Player’s Body-Orientation to Model Pass Feasibility in Soccer + More analysis
• Multi-Person tracking in Basketball
• Self-Supervised Small Soccer Player Detection and Tracking  +  code
• Soccer Action Recognition
• Role classification of soccer players   +   Project page
• Soccer video summarization and highlights  +  Project page

​​​

Non-local methods  (slides of a related talk) 

• Non-local data comparison : axiomatic approach  +  practical approximation
• Image and video denoising
• Exemplar-based inpainting
• Patch-based image segmentation
• Data Comparison,  online demo and code
• Video comparison
• Spatio-temporal video segmentation

​​​

Automatic video analysis and processing 

• Joint occlusions and motion estimation   +   online demo and code  +   more analysis
• General optimization strategy for large displacement variational optical flow +   more analysis, online demo and code
• Depth estimation
• Joint Scene flow and occlusion estimation
• Motion inpainting  +   more analysis, online demo and code
• Video editing: variational dynamic shapes disocclusion
• Slow camera motion   +  demo

​​​

Computational photography 

• Flare artifact detection and removal  +  code
• De-interlacing

​

​​​

Visual perception 

• On the most likely interpretation of the 3D scene structure from a planar image
• Monocular Depth Ordering using Perceptual Occlusion Cues

​​​

Mathematical morphology  

• The tree of shapes of an image
• Contrast invariant image intersection
• Level Lines Selection with Variational Models for Segmentation and Encoding
• The M-components of level sets of continuous functions in WBV

​​​

Mathematical analysis of image processing 

• A Variational Model for P+XS Image Fusion
• Filling-in, inpainting or disocclusion
• Disocclusion by joint interpolation of vector fields and gray levels
• Shape from texture and texture segmentation by variational methods
• Affine invariant image segmentation by variational methods
• Lower Semicontinuity of the Affine Total Variation
• Morse description and morphological encoding of continuous data
• A TV Based Restoration Model with Local Constraints

​​​

The TV flow 

• Minimizing Total Variation flow
• The Dirichlet Problem for the Total Variation Flow

​

​​​

Object detection and recognition

Former and present PhD students

Intelligent Multimodal Vision Analysis Group

Link to some of our reserch projects on github and on hugging face. More software and demos also here.