Julien Hongsavanh

Image processing | Computer vision

Moves to Toulouse

  • 43.6043
  • 1.4419
  • Indicative rate €180 / day
  • Experience 0-2 years
  • Response rate 100%
  • Response time 2h
Propose a project The project will only begin when you accept Julien's quote.

Availability not confirmed

Propose a project The project will only begin when you accept Julien's quote.

Location and geographical scope

Location
Toulouse, France
Can work in your office at
  • Toulouse and 50km around

Preferences

Project length
  • ≤ 1 week
  • ≤ 1 month
  • Between 1-3 months
Business sector
  • Aviation & Aerospace
  • Food Industry
  • Digital Agency and IT company
  • Architecture & Urban Planning
  • Arts & Crafts
+7 autres
Company size
  • 1 person
  • 2 - 10 people
  • 11 - 49 people
  • 50 - 249 people
  • 250 - 999 people
+2 autres

Verifications

Influence

Github

Github : suchiz suchiz
  • 3 Followers
  • 11 Repos
  • 0 Gists

Languages

  • Français

    Native or bilingual

  • Anglais

    Full professional proficiency

Categories

Skills (26)

Julien in a few words

Fraichement diplôme d'un master Informatique Graphique et Analyse d'Images, je me lance en Freelance. J'ai une année d'expérience en vision par ordinateur. Où j'ai entrainer un modèle Mask-RCNN pour les défauts de surface. J'ai une petite expérience en recherche dans la vision par ordinateur toujours, où je vais détecter et reconnaitre des objets 3D dans des nuages de points générée par un LiDar. J'ai toujours été passionné d'informatique. La programmation est quelque chose qui fait partie de mes hobbies. J'ai fait beaucoup de hackathon pour ces même raisons.

Experience

Safran Engineering Services

Image Processing and Computer vision Apprentice

Région de Toulouse, France

September 2019 - Today

1st project: Adaptation of MaskRCNN for surface defects segmentation then adaptation for everything (generic code with only few parameters). I did the whole training and added an evaluation tool. I also added some metrics such as instance callback, and k fold cross validation to ease the training. I'm currently at 0.85 precision and 0.76 recall with a 800 images dataset. We are aiming for 0.90. I also developed a server flask (web platform without interface) to communicate with the android app. The android application is made from scratch, it just sends a picture from your device to the server, and the server sends back the results of the NN. But the app is user friendly, there's a lot of side tools such as gallery and sharing. 2nd project: Implementing a software for automatic electric diagram understanding. -Implementation of line tracker -Basic image processing to cleanse the image (filters, binarisation, skeleton ...) -Text detection and recognition by using Tesseract -Training of MaskRCNN to detect the region of interest -Training of MaskRCNN to detect arrows (Python/OpenCV/PyTesseract/Keras)

UQAM | Université du Québec à Montréal

Infography Laboratory Intern

Montréal, Québec

April 2017 - July 2017

Add a translation option to the multibrush tool in Krita (C++/Qt). Developing games on Unity (C#). Implementing shadowmaps to Gamnit 3D API (Nit/OpenGL ES 2.0).

Education

charter modal image

Success is a team effort

Contribute to this success and the community's professionalism by signing the Freelancer Code of conduct

Sign the code