Accurate re-identification (ReID) of individual insects is crucial for quantitative studies of pollinator behavior, with key applications in biological research and ecological monitoring. This work leverages vision foundation models to enable the fine-grained ReID of honey bees from video with a single paint marking and a single reference track. Such marking avoids the disruption required for gluing tags or painting color codes. We present a new challenging dataset of 9495 images and 45 identities, obtained at outdoor bee feeders with significant pose and illumination changes.Our one-shot capable approach first pre-processes video footage to extract pose-normalized bee crops and remove the background using a segmentation foundation model (e.g., SAM2). It then uses a self-supervised visual foundation model (e.g., DINOv3) for image and patch embeddings, coupled with contrastive metric learning and track information to generate robust ID embeddings for ReID. Compared to existing methods, our approach significantly reduces training data requirements. Our extensive studies show how the choices of the different steps of the pipeline impact performance, offering practical insights for future animal re-identification.

International audience

In this paper, the recognition of pollen bearing honey bees from videos of the entrance of the hive is presented. This computer vision task is a key component for the automatic monitoring of honeybees in order to obtain large scale data of their foraging behavior and task specialization. Several approaches are considered for this task, including baseline classifiers, shallow Convolutional Neural Networks, and deeper networks from the literature. The experimental comparison is based on a new dataset of images of honeybees that was manually annotated for the presence of pollen. The proposed approach, based on Convolutional Neural Networks is shown to outperform the other approaches in terms of accuracy. Detailed analysis of the results and the influence of the architectural parameters, such as the impact of dedicated color based data augmentation, provide insights into how to apply the approach to the target application.

This paper deals with the trajectory estimation of a monocular calibrated camera evolving in a large unknown environment, also known as monocular visual simultaneous localization and mapping. The contribution of this paper is threefold: 1) We develop a new formalism that builds upon the so called Known Rotation Problem to robustly estimate submaps (parts of the camera trajectory and the unknown environment). 2) In order to obtain a globally consistent map (up to a scale factor), we propose a novel loopy belief propagation algorithm that is able to efficiently align a large number of submaps. Our approach builds a graph of relative 3D similarities (computed between the submaps) and estimates the global 3D similarities by passing messages through a super graph until convergence. 3) To render the whole framework more robust, we also propose a simple and efficient outlier removal algorithm that detects outliers in the graph of relative 3D similarities. We extensively demonstrate, on the TUM and KITTI benchmarks as well as on other challenging video sequences, that the proposed method outperforms the state of the art algorithms.

In this paper, we are interested in estimating global motions (homographies, 3D rotations, 3D Euclidean motions, etc.) as well as the covariance of the estimation errors from relative measurements by exploiting the Lie group structure of the motions. We propose a generative model based on the formulation of a concentrated Gaussian distribution on matrix Lie groups. In this context, the global motion estimation problem reduces to the minimization of the sum of squared intrinsic invariant (w.r.t the right action of the Lie group on itself) errors. We derive an iterated extended Kalman filter on matrix Lie groups from the Gauss-Newton formalism on matrix Lie groups, which exhibits a low computational complexity. Experimental results on simulated data, in the context of a consistent pose registration problem, show that the proposed algorithm significantly outperforms the state of the art approaches.

In this paper we address the problem of location recognition from visual lifelogs by leveraging visual features and temporal information in an unified framework. The proposed method features a co-training approach that takes advantage of both labeled and unlabeled data using a confidence measure we propose for this task. It exploits jointly two SVM classifiers on two types of visual features as well as the temporal continuity of the video through temporal accumulation scheme. We demonstrate experimentally on the publicly available IDOL2 dataset that the algorithm yields performance improvement due to its ability to exploit jointly multiple cues, time and unlabeled data.

Our research focuses on analysing human activities according to a known behaviorist scenario, in case of noisy and high dimensional collected data. The data come from the monitoring of patients with dementia diseases by wearable cameras. We define a structural model of video recordings based on a Hidden Markov Model. New spatio-temporal features, color features and localization features are proposed as observations. First results in recognition of activities are promising.

In this paper, we describe a new application for multimedia indexing, using a system that monitors the instrumental activities of daily living to assess the cognitive decline caused by dementia. The system is composed of a wearable camera device designed to capture audio and video data of the instrumental activities of a patient, which is leveraged with multimedia indexing techniques in order to allow medical specialists to analyze several hour long observation shots efficiently.

In this paper, a new formulation based on bidirectional composition on Lie groups (BCL) for parametric gradient-based image alignment is presented. Contrary to the conventional approaches, the BCL method takes advantage of the gradients of both template and current image without combining them a priori. Based on this bidirectional formulation, two methods are proposed and their relationship with state-of-the-art gradient based approaches is fully discussed. The first one, i.e., the BCL method, relies on the compositional framework to provide the minimization of the compensated error with respect to an augmented parameter vector. The second one, the projected BCL (PBCL), corresponds to a close approximation of the BCL approach. A comparative study is carried out dealing with computational complexity, convergence rate and frequence of convergence. Numerical experiments using a conventional benchmark show the performance improvement especially for asymmetric levels of noise, which is also discussed from a theoretical point of view.

A new method for template-based image alignment is presented in this paper. A gradient-based optimization of motion compensated error difference is addressed to solve the image alignment problem. Its novelty lies in the minimized error, which considers the bi-directional compensation of the reference template and the current image onto each other in a constrained asymmetrical fashion. The proposed approach is shown to be a generalization of previous state-of-the art gradient-based methods. An experimental evaluation is provided to show how the new method outperforms the former in presence of noisy images, and to give some insights into its properties.

7-9 novembre 2005

A framework is presented that produces the mosaic corresponding to the background object of an image sequence. It is based on the dominant motion assumption, which states that the background has a parametric motion and occupies the main part of the images. The foreground objects are localised by their different motion. This localisation is computed together with the background motion in an iterative method. The regions corresponding to the background are then pasted onto the mosaic using classic methods adapted to object elimination or a new mosaicking method based on a striping that takes the foreground objects localisation into account.