In this video, we will see Mask R-CNN Project ideas such as verifying the defects of an object and improving and increasing production with this powerful algorithm. We often see Mask R-CNN used to identify people, animals, bicycles, and cars and it does very well but today we will see how to apply this algorithm to projects for practical use and for the industry.

First, we will see why and when we need to use Mask R-CNN instead of other projects and then some project ideas from research.

Why Mask R-CNN Project instead of another algorithm

Before going ahead let’s quickly see what kind of computer vision projects can be realized and then we will see some Mask R-CNN projects

1. Object classification
The algorithm finds the object in the photo but gives no further information

object classification

2. Object detection
The object is identified and the precise position is obtained, even if there are more objects it is possible to distinguish them. An example of an algorithm that allows us to obtain these results is YOLO

object detection

3. Semantic segmentation
Detect the boundaries of the object, creating the classification of each pixel of the identified object but cannot understand if they are one or more objects. An example is U-NET convolutional networks for biomedical image segmentation.

semantic segmentation

4. Instance segmentation
Detect the boundaries of the object and perfectly identifies objects. The perfect example of this algorithm is Mask R-CNN.

Mask R-CNN Project instance segmentation

Here are some interesting Mask R-CNN Project ideas

I have found some interesting projects made with Mask R-CNN and I will list some of them, with links to the research so that you can analyze them and verify the authors. The aim is to present practical applications with clear industrial use instead of the usual (albeit excellent) examples of this interesting algorithm.

Strawberry Diseases

The aim of the project is to identify strawberry diseases with low cost and good accuracy.
Read more about the research: An Instance Segmentation Model for Strawberry Diseases
Based on Mask R-CNN

Plant counting

Count the plants, calculate the homogeneity, and see the size. This is what it is possible to do through a drone and have the images processed by a deep learning algorithm like Mask R-Cnn. I made something like counting crops in a broccoli plantation but with only object detection.
Read more about the research: Mask R-CNN Refitting Strategy for Plant Counting and Sizing in UAV Imagery

Waterline detection

Another interesting application can be in sports. Usually, with computer vision, pose estimation of the person is made to identify the best technique, in this case not. In the sport of canoeing, it is also important to analyze the waterline.
Read more about the research: Utilizing Mask R-CNN for Waterline Detection in Canoe Sprint Video Analysis

Identify the water vapor of the plumes

With other tools, it is difficult to identify the water vapor due to the background but with the right training, this result can also be achieved.
Read more about the research: Utilizing Mask R-CNN for automated segmentation of condensed water vapor plumes from multi-view imagery