The objective of the project is to provide an evidence of concept: Can weuse a system of recognition of images to help in the diagnosis of drowning ?

Clinical issue : accelerate and simplify the work of technicians

Technical issue : automate the process of recognition of diatoms



Link to our final report (in french)
Link to our video (in french)



Group :

Tutor :

  • Bernard Bayle (Teacher at Telecom Physique Strasbourg.)

Clients :

  • Jean-Sebastien Raul (Teacher, Medical school, Forensic Medecine Institute)
  • Evelyne Jehl(Technician, Forensic Medecine Institute)

Manager :

  • Nicolas Padoy (Professor, iCube Laboratory)

Présentation


The diagnosis of drowning is made by the detection of diatoms.

These diatoms are unicellular bodies which resist decomposition and which so find themselves in fabrics after the death.

The first tool of our project is a large number of images of diatoms,(which form our database).We got them at the Forensic Medecine Institute.

The purpose of the project is to use this database to establish a model allowing then the software to recognize a diatom in an image which we supply it..

The project will have to answer specifications whose reliability is one of the key points : false negatives are unacceptable in view of the use of the data in a legal framework.


Steps of the project