Student Group Work

Student Group Work#

Overview#

Your turn! The goal of this group work is to apply the knowledge you have acquired in the course to develop a cell (instance) segmentation pipeline on 2D images provided. Think of using all the tools you have learned in the course: smoothing, thresholding, morphological operations, thresholding, watershed, deep learning, fine-tuning, etc. The only constraint is that your predicted masks should be .png images of the same size as the ground truth.

Segmentation Challenge

Challenges#

Segment the cell outlines.

You are given a training set of 7 images and their corresponding ground truth masks. Your task is to take advantage of this training set to develop a segmentation pipeline that can segment the cell outlines in the images as closely as possible to the ground truth.

Download the dataset#

Download the segmentation challenge dataset

Solution#

You can find the solutions and run it using this Google Colab notebook:

Open in Colab Open in Colab

Evaluation metrics#

Evaluating segmentation pipelines is a complex task, and there is no single metric that can capture all the aspects of the performance. We have chosen to use the following metrics:

  • mAP: The mean average precision is the average of the average precision for each class. It is computed as the area under the precision-recall curve.

\[ \text{mAP} = \frac{1}{n} \sum_{i=1}^{n} \text{AP}_i = \frac{1}{n} \sum_{i=1}^{n} \int_{0}^{1} \text{Precision}_i(r) \text{Recall}_i(r) dr \]
Contents