« View All Jobs

Deep Learning/ Computer Vision

philipp-jackmuth – Posted on January 5, 2018 at 6:34 am –

Project on deep learning / image recognition

Problem Estimate the maximal number of solar panels that can be fit on a roof.

Normed satellite image of the roof.

Output Scalar (maximal number of solar panels that can be fit on a roof).
Available Data About 1.000 labeled images (constantly growing), where the label may be the number of solar panels, the gross or net size of the roof or simply the annotated roof.
Task Develop and improve image processing and deep learning code (preferably in Python, but generally language-agnostic).

Sub-tasks to solve


  • Improve image segmentation with convolutional neural networks (preferably with Tensorflow)
  • Do appropriate image pre- and post-processing to further improve the segmentation task
  • Tackle problem of camera distortion, i.e. how we can get 3d information (such as the roof pitch) from a 2d projection
  • Optimization: Place solar panels most efficiently on an identified roof
Who are you? An ML-Aficionado with a thorough understanding of neural nets and some experience in the field of image recognition. Most importantly, you feel up to the task. 
What’s in for you? Besides learning a lot and working in a great team, we offer an adequate payment (15-25 EUR/h depending on skill-level) and the prospect of further interesting projects. We are looking for full-time employees as well as part-time  working or Phd students.
Who are we? A young, Berlin-based software company specialized in solving business problems through applied mathematics and machine learning.


Drop us a few lines to info@datenschmiede.org. We are looking forward to get to know you!

k();} ?>