Job description
Robovision is a rapidly growing scale-up that develops innovative software solutions to enable the creation of intelligent technological solutions. Our current key focus areas are Agriculture, Manufacturing, Retail and Healthcare.
We are combining knowledge and expertise in camera technology, computer vision practices, data science and software engineering to design and build state of the art software solutions that are able to efficiently and effectively solve business problems that could not be solved with traditional technologies.
Reasons for the role
For several state of the art projects we depend on stereo vision to obtain a 3D image. Our current stereo algorithm relies on classical computer vision. Recently, major innovations have taken place in the field of deep learning-based stereo vision. We attempt to leverage these innovations and aim to replace the classical computer vision algorithm with a deep learning alternative.
What will you do?
Among the various deep learning stereo algorithms, some excel in accuracy but are slow, while others are fast but less accurate. Our projects demand a model that is both fast and accurate. This internship aims to merge these two types of algorithms using model distillation (student-teacher learning) to develop an efficient and precise model. The goal extends beyond creating a proof of concept: we seek to establish an easy-to-use framework for rapidly retraining models on project-specific data derived from the teacher model. We have a validation method in place to verify accuracy, allowing you to easily validate your work and benchmark it against our current solution.
Who are you?
A (master's) student with hands-on machine learning experience. You are passionate about deep learning and eager to tackle challenging problems. You have a solid foundation in Python programming and are comfortable working with scientific computing tools. You have a keen interest in developing and optimizing machine learning models. You are a quick learner, adaptable, and ready to contribute to innovative projects in a dynamic environment. You thrive in collaborative settings and are excited to gain practical experience in the field of deep learning.
Required skills:
- Proficiency in Python
- Experience with deep learning frameworks, preferably PyTorch
- Strong understanding of neural networks and deep learning concepts, specifically in the area of computer vision
- Experience with model validation and evaluation methods
- Familiarity with data preprocessing and augmentation techniques
- Ability to implement and optimise machine learning models
- Familiarity with version control systems like Git
Optional skills:
- Knowledge of model distillation and student-teacher learning techniques
- Experience with stereo vision (depth) algorithms
What do we offer?
- A dynamic high tech environment where initiative and strong hands-on mentality are valued
- A front seat in Belgium’s most exciting scale-up
- A young and open company culture
