Open postdoc position

A fully funded postdoctoral position in machine/deep learning for computational microscopy opens in 2019 at Telecom SudParis for a collaborative project:

Reconstruction and detection from sparse Fourier ptychographic microscopy

Label free imaging provides data from biological specimens without modifying them, which is a step forward in tracking physio-pathological mechanisms. In particular, Fourier ptychographic microscopy (FPM) proposed in 2013 reconstructs high-resolution intensity and phase images from hundreds of acquisitions under controlled lighting incidence [1]. It allows contrasting unstained biological objects and in-depth focusing [2], for example to further address cells of interest and capture unaltered molecular signatures [3]. Telecom SudParis is developing a FPM system with a partner SME.

This project will study the efficient use of FPM to locate blood cells and extract discriminant features in a progressive and parsimonious approach to sensing and processing. A first part of the work will consist in enhancing the stability and accuracy of the intermediate intensity and phase images that are reconstructed during capture [4] [5] or from sparse acquisitions. Then, multimodal detection and characterization of objects of interest will be investigated, in particular using convolutional networks and deep learning [6] [7]. An efficient implementation, in the spectral domain, on the GPU [8] and from partial acquisitions [9] is expected.

Keywords: Machine learning, Deep learning, Computational imaging, Microscopy.

Location: Telecom SudParis, in Evry (near Paris), France.
Telecom SudParis is a leading public graduate school of engineering in Information and Communication Technologies. It is part of Institut Mines-Telecom, France's leading group of engineering schools, and it is a member of Université Paris‑Saclay, the first French research cluster in sciences and technologies of information. The 105 full time professors of Telecom SudParis contribute to the education of 1,000 students including 120 doctoral students.

Duration: 12 months, from January to December 2019.

Contact: Inquiries and applications (cover letter and CV, with recommendation) should be sent to Patrick Horain ( Applications will be continuously received until the position is filled.


[1] G. Zheng, R. Horstmeyer, C. Yang (2013). "Wide-field, high-resolution Fourier ptychographic microscopy". Nature photonics, 7(9), 739-745. [doi:10.1038/nphoton.2013.187] 

[2] L. Tian, L. Waller (2015). "3D intensity and phase imaging from light field measurements in an LED array microscope". Optica 2, 104-111. 

[3] T. Happillon et al. (2015). "Diagnosis approach of chronic lymphocytic leukemia on unstained blood smears using Raman microspectroscopy and supervised classification". Analyst, 140 (13), 4465-4472. 

[4] C. Dogariu, P. Horain (2017). "Progressive on-the-fly Fourier ptychography reconstruction". Face2Phase conference, Delft, Nederland. 

[5] Y. Zhang, W. Jiang, L. Tian, L. Waller, Q. Dai (2015). "Self-learning based Fourier ptychographic microscopy". Optics Express 23 (14), 18471-18486 [doi:10.1364/OE.23.018471]. 

[6] O. Ronneberger, P. Fischer, T. Brox (2015). "U-Net: Convolutional Networks for Biomedical Image Segmentation". Medical Image Computing and Computer-Assisted Intervention (MICCAI), Springer, LNCS 9351, 234--241. [arXiv:1505.04597, doi:10.1007/978-3-319-24574-4_28]. 

[7] J. C. Ye, Y. Han, E. Cha (2018). "Deep convolutional framelets: a general deep learning framework for inverse problems". SIAM Journal on Imaging Sciences 11(2), 991–1048. [arXiv:1707.00372]. 

[8] Y. Allusse, P. Horain et al. (2008). "GpuCV: An Open Source GPU‑Accelerated Framework for Image Processing and Computer Vision". Proceedings of the 16th ACM international conference on Multimedia, 1089-1092, Vancouver, BC, Canada [doi:10.1145/1459359.1459578]. 

[9] R. Horstmeyer, R. Y. Chen, B. Kappes, B. Judkewitz (2017). "Convolutional neural networks that teach microscopes how to image". [arXiv:1709.07223v1].