Engineering

Data Scientist Intern

Paris, Toulouse

Alteia

Alteia is an industrial asset management platform based on artificial intelligence. The company capitalizes on a decade of experience in capturing and analyzing images. Its Visual Intelligence offer thus combines the best of computer vision and AI technologies. The platform ingests and structures very large amounts of field data into a single source of truth, which makes it possible to develop predictive models for industrial infrastructures. It optimizes the entire life cycle of an operation, simplifies risk management and provides real-time information on these infrastructures.

Your Missions

Within our Toulouse offices, you will join a team of data scientists specializing in deep learning, image processing and geomatics, in charge of R&D on data analysis tools deployed on the delair.ai platform. The team’s expertise relates to the analysis of data from a wide variety of sources: images of drones, LIDAR point clouds, CAD models, etc.

3 themes are considered according to your profile:

Deep learning applied to 3D data.

Your main task will be to design algorithms for the semantic segmentation of point clouds.

Point clouds from LIDAR sensors and / or photogrammetric processing generally have large sizes and require robust and efficient algorithms to be analyzed.

First, you will learn about a semantic segmentation model developed internally to familiarize yourself with the problem.

Secondly, you will have to implement and compare different neural network architectures adapted to this type of data: deep nets, 3D CNN, point-net, CNN graph, etc.

Skills valued:

  • Machine Learning / Deep learning concepts.
  • Python.
  • Experience with a machine learning library (PyTorch, Tensorflow, sklearn).

Connections:

  • PointNet ++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space: https://arxiv.org/abs/1706.02413.
  • KPConv: Flexible and Deformable Convolution for Point Clouds: https://arxiv.org/abs/1904.08889.
  • PyTorch Points 3D: https://github.com/nicolas-chaulet/torch-points3d.

Deep learning applied to semantic image segmentation.

Your main task will be to design algorithms for semantic segmentation on orthorectified images from a drone survey of industrial infrastructure (agriculture / mining and quarry).

First, you will learn about a semantic segmentation model developed internally to familiarize yourself with the problem.

As a second step, you will have to implement and test different neural network architectures that use semi-supervised or weakly supervised approaches.

Skills valued:

  • Machine Learning / Deep learning concepts.
  • Python.
  • Experience with a machine learning library (PyTorch, Tensorflow, sklearn).

Connections:

  • Semi-supervised semantic segmentation needs strong, varied perturbations: https://arxiv.org/pdf/1906.01916v4.pdf.
  • Hierarchical multi-scale attention for semantic segmentation: https://arxiv.org/pdf/2005.10821v1.pdf.
  • Invariant Information Clustering for Unsupervised Image Classification and Segmentation: https://arxiv.org/pdf/1807.06653.pdf.

Deep learning applied to object detection on images.

Your main task will be to design algorithms for semantic segmentation on images from a drone survey of industrial infrastructure (power lines).

First, you will learn about object detection models developed in-house to familiarize yourself with the problem.

Secondly, you will have to implement and test different neural network architectures that aim to improve performance through a so-called multi-view approach that effectively uses different points of view on the same object.

Skills valued:

  • Machine Learning / Deep learning concepts.
  • Python.
  • Experience with a machine learning library (PyTorch, Tensorflow, sklearn).

Connections:

  • Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks: https://arxiv.org/abs/1506.01497.
  • D2-Net: A Trainable CNN for Joint Detection and Description of Local Features: https://arxiv.org/abs/1905.03561.
  • End-to-end object detection with Transformers: https://ai.facebook.com/blog/end-to-end-object-detection-with-transformers/.

Your Profile

Student in the last year of a Bac + 5 type Master 2 or Engineering School with a specialization in image processing or Data science.

Apply to this job

Data Scientist Intern







    Alteia is committed to protecting and respecting your privacy, and we’ll only use your personal information to provide the information requested from us.