[OPEN] [ISL_CV_VP_003] Big Data Coding using Machine Learning and Patch Manifolds 

Due to the massive amount of pictures stored by Facebook, YouTube and other applications, there is a need to efficiently compress similar images to reduce storage capacity requirements. The redundancy of those similar images can be successfully exploited by leveraging current image/video coding schemes.

In this internship we would like to investigate new paradigms using machine learning on patch manifolds to predict or restore image patches. In particular we would like to study those techniques as a pre and/or post-processing added to traditional HEVC/JPEG encoders

The candidate will have to implement and evaluate various methods based on manifolds, machine learning, and minimization problems. He/she will use an existing encoding framework for validation. He/she will work with other people working on similar issues.

Skills : Computer vision and image processing, Machine learning, image/video compression, applied mathematics, C/C++ programming

Keywords : Video compression, Machine Learning

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [ISL_CV_HCI_VP_005] Fast semi-automatic rotoscoping for high-end VFX 

Technicolor hosts world’s leading visual effects (VFX) studios, creating high-end VFX for the advertising and feature film industries. The goal of this internship is to study, test and implement state-of-the-art algorithms and tools to help studio artists with the ubiquitous and time-consuming task of rotoscoping (cutting out objects from the background) in video shots. The intern will have the opportunity to work with and to enhance professional tools and frameworks. Check what we do at http://www.moving-picture.com/reels/vfx-breakdowns/.

Skills : Engineer or master student with a solid background in computer vision and machine learning; Strong mathematical background; Good programming (C++, python) and software design skills; Hands-on video compositing software is a plus; Good written/oral communication skills; Fluent in English

Keywords : Computer vision, object tracking, motion analysis, motion/image saliency, user interfaces

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[CLOSED] [ISL_CV_008] Deep Learning based Image analysis on mobile 

Deep Learning technology applied on image analysis and editing has opened up a full range of new applications. Some of them, ranging from image upscaling, image tagging to object recognition, make much sense in the context of mobile phone. The focus of the internship is to study the adaptation and port of deep learning image processing and model for mobile platform.

Skills : Knowledge in Computer Vision and Machine Learning, Ability to read English scientific literature

Keywords : Deep learning, computer vision, embedded environment

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [IML_CV_010] Collaborative Indoor localization using computer vision 

When mobile device users want to know what their localization is, they usually use GPS like sensors. While these sensors are very powerful, they are not compatible with indoor use and do not give information about orientation.

This localization is useful in numerous applications, starting with augmented reality. This internship is part of a wider project on mixed reality with dedicated glasses or tablets. The intern will work with several engineers and scientists specialized in this field.

Another way to localize a device is to use the embedded camera. Using state of the art computer vision methods, some algorithms enable to recognize previously visited places. Indoor environments being visually poor, developed algorithms will have to use as many information as possible.

The internship will start with a complete bibliography on the subject. Then, the intern will work together with a senior scientist to create a localization method which will have a high positive results rate. Then the internship will focus on how to make the measurement made on one device compatible with heterogeneous devices.

This internship requires a strong scientific behaviour and a good level in mathematics (and, if possible, in computer vision). This internship may be the starting point of a CIFRE thesis under supervision of an international grade research team. Student applying to this job must be interested in continuing studies.

Skills : Computer vision, Mathematics, Bayesian, Machine Learning, Deep Learning, Multisensors

Keywords : Place recognition, Localization, Data fusion, Camera, sensors, augmented reality

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [IML_CV_VP_011] Development of an Android Application for Live Events Animation 

The proliferation of smartphones allows access to people’s position and message notifications in real-time. During live events (sports, concerts, spectacles…), this source of information and communication maybe successfully used to provide the organizers with interactive applications, enhancing the users experience.

This internship will thus consist in implementing a client/server application for Android smartphones dedicated to those special events.

The candidate will have to i) implement the terminal app, ii) the server application to manage several clients and send back the appropriate information, and iii) perform the necessary field trials and user evaluations.

He/she will work with other people within our research facilities.

Skills : Computer vision and image processing, Android programming, applied mathematics

Keywords : Android app., Image processing

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[CLOSED] [IML_CG_CV_VP_013] Image style transfer from real scene videos to virtual / mixed scene videos 

Deep learning is demonstrating its applicability to a wide set of computer vision problems and in particular it has shown recently its potential for high level image synthesis and manipulation.

In this context, the objective of the internship is to develop a fast method based on neural networks to transfer style in images of real scenes to synthetic images for applications in augmented reality.

The internship will consist of an analysis of the state of the art in style transfer and deep learning, the specification of a solution suited to our needs, and the implementation, testing and improvement of a method.

Photo : An example of image style transfer

Skills : Specific knowledge in computer vision, good programming skills in C/C++ and Python, Windows, possibly CUDA.

Keywords : Computer vision, deep learning, style transfer, mixed/augmented reality, video processing, real-time rendering.

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [ISL_CV_VP_026] Combining Unsupervised and Supervised learning to grade SDR to HDR videos contents 

High Dynamic Range displays have become more and more popular during the international electronics exhibitions. They offer a new user experience as they can show images and videos with high brightness compared to standard displays. HDR devices are able to display videos with more details in black levels and higher contrasts. Technicolor has developed Intelligent Tone Management (ITM) technology that aims converting video from Standard Dynamic Range (SDR) to High Dynamic Range (HDR) avoiding banding issues, noise amplification and enhance local and global contrast when luminance is extended.

The goal of this internship is to explore and propose methods to grade automatically SDR to HDR images and videos contents. Based on ground-truth graded manually by colorists, the intern has to use machine learning approaches (supervised or/and supervised) to extract a model that can be used to grade a new video content

 

 

 

Skills : Matlab/Python/C programming, ideally with image processing expertise - - Ability to write well-structure and documented code - - Good written and spoken English - Excellent team working skills as the internship forms a part of a larger project, involving many team members - Ability to work independently

Keywords : HDR, Machine Learning, Deep Learning, SVM, Clustering, ITMO (Inverse Tone Mapping)

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[CLOSED] [ISL_CV_HCI_VP_029] Contextual Interestingness of Media Content 

Knowing whether a media content, i.e., image or video, is interesting for a given viewer has numerous applications from assets management, improved education, to targeted advertising.

This internship proposal is a following of already existing reseach activities at Technicolor. It targets the improvement of algorithms for content interestingness prediction and the development and implementation of new algorithms for contextual interestingness prediction. These algorithms will be based on machine learning techniques such as deep learning. One expected output is to submit a system to the 2017 MediaEval (http://www.mutimediaeval.org) task on interestingness.

Skills : machine learning, computer vision, image processing, Python, C++, OpenCV. Audio processing skill would be a plus

Keywords : machine learning (deep learning), image & video processing, multimodal fusion

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [ISL_CV_HCI_VP_030] Predicting Memorability of Media Content 

Knowing whether a media content, i.e., image or video, is memorable for a viewer has numerous applications from assets management, improved education, to targeted advertising.

This internship proposal targets the development and implementation of such a memorability prediction algorithm based on machine learning techniques such as deep learning.

Skills : machine learning, computer vision, image processing, Python, C++ or Java

Keywords : machine learning (deep learning), image & video processing, multimodal fusion

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [IML_CV_VP_036] Multi-View Video Manipulation 

Nowadays, there is a major trend to have devices with multiple cameras in order to improve the image quality and to propose new applications compared to conventional video. While video processing tools are well-known, there is still a real need for developing specifically tailored algorithms in the multi-camera framework such as depth estimation, tracking, segmentation, etc… The goal of this internship is to explore and propose new methods for manipulating and mixing multi-view videos

Skills : Computer vision and image processing, C++ programming

Keywords : Multi-Camera, Light-Field, Video processing

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [HOME_CG_CV_VP_040] Model Reconstruction Using Multiple Depth Sensors 

This internship focuses on real-time reconstruction of 3D models (Point Clouds) in the context of Telepresence scenarios. The trainee will design real-time algorithms on GPU in order to reconstruct clean and stable 3D point clouds from a “non-rigid” rig of depth cameras such as Kinect V2 or RealSense. The work will take place in research project focusing on Immersive technologies.

Skills : C/C++, GPU Programming (GLSL, Compute Shaders), OpenGL, [Python, Java Script]

Keywords : Reconstruction, Depth Cameras, Computer Vision, Augmented Reality, Mixed Reality, Telepresence, Kinect V2, RealSense.

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.

[OPEN] [ISL_CV_VP_041] Deep net with deep loss 

Most state-of-art deep nets are trained in a supervised way, using annotated training data and classic classification or regression losses. Very recent works have showed that, in the context of self-supervised training, a new kind of loss based on another pre-trained network could be used with several appealing properties. This approach has been demonstrated for learning feed-forward neural nets either for non-photorealistic image editing or for image up-sampling

This internship aims at developing further this promising type of approach, using other pre-trained losses and attacking other types of image processing and editing tasks

Skills : machine learning, computer vision, Python/C++

Keywords : deep learning, image processing and editing

 

This internship is located in Rennes, France. If interested, please apply at stage.rennes@technicolor.com by sending us your resume and a cover letter with the internship reference in the email subject line.