Hello Kailan,
I understand that you want to create an R-CNN model that not only detects road defects like potholes and cracks but also assigns a condition score to the detected defects.
You can follow the below steps to build the required R-CNN model:
- Label your dataset: You can use MATLAB's Image Labeler or Video Labeler app to manually label your dataset with the bounding boxes and their corresponding categories (e.g., "pothole" or "crack") and condition scores (1 to 5). Then, divide your dataset into training and validation sets.
- You have to modify the R-CNN architecture to have two separate output layers: A classification layer for detecting the type of defect (e.g., "pothole" or "crack") and a regression layer for predicting the condition score (1 to 5).
- Train and test the R-CNN: Train the modified R-CNN architecture on your labelled dataset using the appropriate training options. Then, use the trained R-CNN to detect defects on new images. The output should include the bounding boxes, defect types, and condition scores for each detected defect.
You can refer to the below documentation to understand more about Object Detection Using Faster R-CNN Deep Learning, Image Labeler and Video Labeler.
