arXiv subjects

Julien Olivier

Publications and source records attributed to Julien Olivier.

5 recordsLinked to original sources

arXiv2024

Automatic Die Studies for Ancient Numismatics

Die studies are fundamental to quantifying ancient monetary production, providing insights into the relationship between coinage, politics, and history. The process requires tedious manual work, which limits the size of the corpora that can be studied. Few works have attempted to automate this task, and none have been properly released and evaluated from a computer vision perspective. We propose a fully automatic approach that introduces several innovations compared to previous methods. We rely on fast and robust local descriptors matching that is set automatically. Second, the core of our proposal is a clustering-based approach that uses an intrinsic metric (that does not need the ground truth labels) to determine its critical hyper-parameters. We validate the approach on two corpora of Greek coins, propose an automatic implementation and evaluation of previous baselines, and show that our approach significantly outperforms them.

cs.CV↗

arXiv2019

From Patch to Image Segmentation using Fully Convolutional Networks -- Application to Retinal Images

Deep learning based models, generally, require a large number of samples for appropriate training, a requirement that is difficult to satisfy in the medical field. This issue can usually be avoided with a proper initialization of the weights. On the task of medical image segmentation in general, two techniques are oftentimes employed to tackle the training of a deep network $f_T$. The first one consists in reusing some weights of a network $f_S$ pre-trained on a large scale database ($e.g.$ ImageNet). This procedure, also known as $transfer$ $learning$, happens to reduce the flexibility when it comes to new network design since $f_T$ is constrained to match some parts of $f_S$. The second commonly used technique consists in working on image patches to benefit from the large number of available patches. This paper brings together these two techniques and propose to train $arbitrarily$ $designed$ $networks$ that segment an image in one forward pass, with a focus on relatively small databases. An experimental work have been carried out on the tasks of retinal blood vessel segmentation and the optic disc one, using four publicly available databases. Furthermore, three types of network are considered, going from a very light weighted network to a densely connected one. The final results show the efficiency of the proposed framework along with state of the art results on all the databases.

cs.CV↗

arXiv2016

Aging and linear response in the Hébraud-Lequeux model for amorphous rheology

We analyse the aging dynamics of the Hébraud-Lequeux model, a self-consistent stochastic model for the evolution of local stress in an amorphous material. We show that the model exhibits initial-condition dependent freezing: the stress diffusion constant decays with time as $D\sim 1/t^2$ during aging so that the cumulative amount of memory that can be erased, which is given by the time integral of $D(t)$, is finite. Accordingly the shear stress relaxation function, which we determine in the long-time regime, only decays to a plateau and becomes progressively elastic as the system ages. The frequency-dependent shear modulus exhibits a corresponding overall decay of the dissipative part with system age, while the characteristic relaxation times scale linearly with age as expected.

cond-mat.soft↗

arXiv2016

Emergence of coherent localized structures in shear deformations of temperature dependent fluids

Shear localization occurs in various instances of material instability in solid mechanics and is typically associated with Hadamard-instability for an underlying model. While Hadamard instability indicates the catastrophic growth of oscillations around a mean state, it does not by itself explain the formation of coherent structures typically observed in localization. The latter is a nonlinear effect and its analysis is the main objective of this article. We consider a model that captures the main mechanisms observed in high strain-rate deformation of metals, and describes shear motions of temperature dependent non-Newtonian fluids. For a special dependence of the viscosity on the temperature, we carry out a linearized stability analysis around a base state of uniform shearing solutions, and quantitatively assess the effects of the various mechanisms affecting the problem: thermal softening, momentum diffusion and thermal diffusion. Then, we turn to the nonlinear model, and construct localized states - in the form of similarity solutions - that emerge as coherent structures in the localization process. This justifies a scenario for localization that is proposed on the basis of asymptotic analysis in \cite{KT}.

math.AP↗

arXiv2015

Probing relevant ingredients in mean-field approaches for the athermal rheology of yield stress materials

Although the notion of mechanical noise is expected to play a key role in the non-linear rheology of athermally sheared amorphous systems, its characterization has so far remained elusive. Here, we show using molecular dynamic simulations that in spite of the presence of strong spatio-temporal correlations in the system, the local stress exhibits normal diffusion under the effect of the mechanical noise in the finite driving regime. The diffusion constant appears to be proportional to the mean plastic activity. Our data suggests that the corresponding proportionality constant is density independent, and can be directly related to the specific form of the rheological flow curve, pointing the way to a generic way of modeling mechanical noise in mean-field equations.

cond-mat.soft↗