TBD

TBD

We developed a pipeline in which the steps taken by the approach are mainly derived from traditional computer vision techniques:

• Changes in color space;
• Histogram equalization;
• Thresholds;
• Morphological operations;
• Feature extraction (using HOG);
• Supervised Learning (with Support Vector Machine, i.e. SVC)

First, you need to build all solution (i.e. train the SVC model) by running the Jupyter Notebook _pipeline.ipynb available at TF/computer_vision/segment_cell. For visualizing each step of the solution, open the Jupyter Notebook source.ipynb available in the TF/computer_vision folder.

• This Part of the TF was developed using the Pytorch framework.
• Regarding the CNN model used in this work, we choose a modifying U-NET architecture where we change its backbone to use a VGG model and we have also included a ASPP module at the bottom of the U-Net model to handle multi-scale segmentations tasks.

First of all, you need to install the NVIDIA Automatic Mixed Precision (Amp). You can find more information about it and also how to install it in this GitHub repo: https://github.com/NVIDIA/apex

To train the model for bacteria segmentation, you have to open the train.py script and modify the paths according to your environment setup. Having this first step done, just open the terminal inside the TF/deep_learning/src folder and type:

python3 train.py

If you just want to test our model using your dataset (considering that your data is related to darkfield microscopy images, see https://www.kaggle.com/longnguyen2306/bacteria-detection-with-darkfield-microscopy for more details), you have to download the pretrained model, available at: https://www.dropbox.com/s/0of9ref6fosit5a/unet_vgg_aspp.pth?dl=0

Otherwise, if you have trained a new model, just follow the steps bellow.

After this, you have to open the test or inference script, both available in the TF/deep_learning/src folder, and again modify the paths according to your environment setup. Then, just open the terminal (or your IDE) and type:

• For Testing:

python3 test.py

• For Inference:

python3 inference.py