The goal is to segment instances of microvascular structures, including capillaries, arterioles, and venules, to automate the segmentation of microvasculature structures to improve researchers' understanding of how the blood vessels are arranged in human tissues.
- Description: API for predicting using the UNet model.
- Input: Kidney tissue image of shape
512x512x3
, and its corresponding mask of shape512x512x1
. - Output:
- 6 paths for the
image
,overlaid_image_true
,overlaid_image_pred
,true_mask
,predicted_mask
, andoverlaid_mask
. - 2 score metrics to indicate confidence
IOU
, andDice
. - Finally, the
threshold
at which thepredicted_mask
was binarized.
- 6 paths for the
- Description: API for predicting using LinkNet model.
- Input: Kidney tissue image of shape
512x512x3
, and its corresponding mask of shape512x512x1
. - Output:
- 6 paths for the
image
,overlaid_image_true
,overlaid_image_pred
,true_mask
,predicted_mask
, andoverlaid_mask
. - 2 score metrics to indicate confidence
IOU
, andDice
. - Finally, the
threshold
at which thepredicted_mask
was binarized.
- 6 paths for the
- Description: API for predicting using FCN model.
- Input: Kidney tissue image of shape
512x512x3
, and its corresponding mask of shape512x512x1
. - Output:
- 6 paths for the
image
,overlaid_image_true
,overlaid_image_pred
,true_mask
,predicted_mask
, andoverlaid_mask
. - 2 score metrics to indicate confidence
IOU
, andDice
. - Finally, the
threshold
at which thepredicted_mask
was binarized.
- 6 paths for the
- Description: API for predicting using an ensemble of models.
- Input: Kidney tissue image of shape
512x512x3
, and its corresponding mask of shape512x512x1
. - Output:
- 6 paths for the
image
,overlaid_image_true
,overlaid_image_pred
,true_mask
,predicted_mask
, andoverlaid_mask
. - 2 score metrics to indicate confidence
IOU
, andDice
. - Finally, the
threshold
at which thepredicted_mask
was binarized.
- 6 paths for the
To install the project dependencies, run the following command:
pip install -r requirements.txt
This command will install all the necessary packages listed in the requirements.txt
file.
To ensure the project runs smoothly, please follow these steps:
-
Weights Directory: Create a directory named
weights
within the model directory. This directory should contain the weights of the three models (UNet, LinkNet, FCN). These weights are essential for the proper functioning of the project. Adjust the model weights file name to match the one in themodel_weights.json
file. -
Images Directory in Static: Create a directory named
images
within thestatic
directory. This directory will contain the images generated by the server to display them later.
Make sure to create these directories and add the necessary files before running the project.