For any information regarding this project, please contact antoine.meilliez@hec.edu
This project aims to detect automatically molecule and transcriptions site involved in the first step of developpement of drosophilia embryo. Such information can be extracted from microscop images such as this one:
This package implements several tools requested to achieve this classification:
- Image filtering and scaling
- Region extraction
- Blob detection
- Contours detection
- 3D region aggregation
- Region features extraction
- Unsupervised classification benchmarker
- File manager and viewer
- Performance monitoring
A standard pipeline has been implemented in demo.py
Region detection performed on cells.
This detection is used as feature of classification algorithm. Region are tagged regarding the cell they are embedded in or not.
Region intensity exploration along the z axis
Intensity / area ratio
Distribution filtering. This part enables classification of Transcription site
Features distribution analysis
A benchmark tool has been implemented to identify which classification algorithm presents the best results.
The following classification was handled on the raw features (prior to distribution filtering)
