Scientific direction Development of key enabling technologies
Transfer of knowledge to industry

PhD : selection by topics

Panoptic segmentation with deep semi-supervised learning

Département Intelligence Ambiante et Systèmes Interactifs (LIST)

Vision & Ingénierie des Contenus (SAC)

01-10-2019

SL-DRT-19-1009

florian.chabot@cea.fr

This thesis focuses on panoptic segmentation which has two main objectives : associate each image pixel with a class (person, car, tree?) and provide a mask for each object of the scene. In other words, this new kind of segmentation merges semantic segmentation and instance segmentation to provide fine description of the image. The first objective of the thesis is to propose a new method for panoptic segmentation using deep supervised learning. The second objective is to extend this method using a smaller training dataset as well as weakly labelled or unlabelled data (semi-supervised learning). The idea behind is to succeed to learn an efficient model for panoptic segmentation using a small set of fully annotated data because the annotation process is time consuming.

Circuit / substrate co-design for 5G Front-End-Module RF applications

Département Composants Silicium (LETI)

Laboratoire Intégration et Transfert de Film

01-10-2019

SL-DRT-19-1021

emmanuel.augendre@cea.fr

The Front End Module (FEM) represents a significant part of a smart phone cost (>3x the application processor). Today, each new smart phone generation contains an increasing FEM area fabricated with RFSOI technology (20x between 2010 and 2016), a dedicated device technology on a dedicated substrate, where STMicroelectronics and Soitec offer state-of-the-art products. Soitec substrates already allow a strong second harmonic attenuation on passive structures. Technology solutions exist to provide: - a higher degree of linearity - temperature-independent linearity - intermediate linearity at lower cost. These solutions have not yet been industrialized. The purpose of this PhD work is to: - optimize the design of one or more FEM functional blocks using state-of-the-art PDSOI technology, - have these circuits fabricated - measure the circuit performance and evaluate the benefit of exploratory substrates.

Online characterization and classification of radiological signals using embedded machine learning

DM2I (LIST)

Laboratoire Capteurs et Architectures Electroniques

01-10-2019

SL-DRT-19-1033

gwenole.corre@cea.fr

Machine learning classification methods have become ubiquitous in the fields of signal and image processing. Nevertheless, these classification methods remain very little used today in the fields of embedded applications. In fact, several studies have shown that machine learning classification methods provide satisfactory performance in classifying signals received from radiological sensors. However, most of these studies and solutions have been developed and experimented offline, and there are few or no real-time neural network learning based solutions. The aim of the thesis is to propose learning methods that can be embedded in real-time portable measurement devices. The proposed methods have to meet the radiological signal classification constraints and to enhance the performance of these measurement devices.

Atomic-level control over ultrathin 2D layers of Transition Metal Dichalcogenides obtained by a Molecular Layer Deposition route

Département technologies silicium (LETI)

Autre laboratoire

01-10-2019

SL-DRT-19-1048

denis.rouchon@cea.fr

2D-materials, especially transition metal dichalcogenides (TMD), have recently received considerable attention since they are emerging as a class of exceptional semiconductor materials with many potential applications (supercapacitors, batteries, electronics and opto-electronics, flexible electronics ...). However, a sizeable bottleneck for their full deployment stems from the lack of scalable fabrication methods with atomic scale precision. Both Atomic Layer Deposition (ALD) and Molecular Layer Deposition (MLD) are based on sequential, self-limiting surface reactions that allow conformal film growth with precise thickness control. They are ideal techniques for depositing scalable ultrathin inorganic and organic films. The thesis project aims to achieve the synthesis of intermediate layered metal-organic hybrid films by a combination of Atomic Layer Deposition and Molecular Layer Deposition (ALD/MLD) followed by mild thermal treatment (annealing) en route to the crystalline final phase. The targeted materials are first TiS2 then SnS2. We want to explore with the thermal treatment a possible route to synthesize graphitic interfaces which could allow new electrodes or electrical contacts. Provide an atomic-level insight and control over the growth by combining ab initio calculations (not to be performed by the PhD student) and in situ chemical and structural studies performed during the growth and thermal treatment, i.e. in situ X-ray absorption and scattering with custom-built equipment, in situ Raman scattering, residual gas analysis, ellipsometry

Optical laser waveguides III-V (AsGa/InP) growth directly onto SOI-300mm.

Département d'Optronique (LETI)

Laboratoire d'integration technologique pour la photonique

01-10-2019

SL-DRT-19-1055

christophe.jany@cea.fr

For more than 25 years, the heart of telecom networks has become one of the fields of application of the III-V components (InP_like, and AsGa_like). This field is based on the transmission of IR waves in optical fibers, powered by laser sources in III-V materials. Over the past ten years, a new technological path has been developed, based Silicon-Photonics, which makes it possible to lower manufacturing costs by increasing integration (3D integration, Wafer Level Packaging). The approach usually chosen here consists of a molecular bonding of a III-V wafer (epitaxial) on an SOI previously structured optical guides. A technological treatment is then applied to make III-V transmitter guides connected to the silicon guides. Since less than 5 years; a new integration scheme is developing, it is the direct epitaxy of III-V materials on silicon. For 3 years, the CEA / LETI laboratories, already experts in the development of photonics on Silicon by bonding process, have decided to investigate this highly innovative approach with high potential. The proposed thesis will thus rely heavily on the CNRS / LTM laboratory, which has been developing for the last 4 years new MOCVD epitaxial concepts for III-V materials (AsGa base) on textured silicon wafer. This subject of study will enable the establishment of a new roadmap of III-V epitaxy on Silicon, with the aim of designing a new generation of photonic circuits. The PhD student will be involved both in the development of III-V materials on silicon, as well as in the design and realization of photonic circuits 2.0.

Multi-point measurement method development for uncertainties reduction of activity quantification in nuclear waste drum using gamma spectrometry

DM2I (LIST)

Laboratoire Capteurs et Architectures Electroniques

01-10-2019

SL-DRT-19-1069

adrien.sari@cea.fr

The management of radioactive waste packages is a major challenge for the nuclear industry. Characterization of the packages requires non-destructive nuclear measurement solutions in order to preserve the integrity of packages. The present thesis will focus on a concrete case of application which will consist in equipping with embedded sensors a six-axis robotic arm bearing a smear system, a contaminameter, and a gamma spectrometer. The robotic arm will allow multipoint measurements, in dose rate (with a Geiger-Müller type detector) and gamma spectrometry (with a CdZnTe detector). The drums to be characterized will have weak or medium activities, and the radioelements to be identified will be activation products and actinides. The aim of this thesis is to define a dynamic multipoint measurement method for optimizing the declaration of uncertainty associated with the quantity of interest (dose rate and activity). This thesis will include an MCNP6 simulation component and an experimental component.

61 Results found (Page 10 of 11)
first   previous  7 - 8 - 9 - 10 - 11  next   last

Voir toutes nos offres