EXPLAINABLE ARTIFICIAL INTELLIGENCE (XAI) USING GRAD-CAM AND EIGEN-CAM ON YOLOV5S MODELS
This thesis discusses the application of Explainable Artificial Intelligence (XAI) in object detection using the You Only Look Once (YOLO) algorithm version 5. Along with the development of artificial intelligence, which is growing rapidly and becoming increasingly complicated, XAI is needed to be a...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/74725 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | This thesis discusses the application of Explainable Artificial Intelligence (XAI) in object detection using the You Only Look Once (YOLO) algorithm version 5. Along with the development of artificial intelligence, which is growing rapidly and becoming increasingly complicated, XAI is needed to be able to provide transparent and understandable explanations to users. This thesis uses the XAI algorithm in object detection using Gradient-weighted Class Activation maps (Grad-CAM) and Eigen-CAM.
The initial stage is to build data sets on several roads in the city of Bandung. Once the dataset is labeled, the YOLOv5s model is trained using the collected dataset. The results of the YOLOv5s model that has been trained get a mean Average Precision (mAP) value of 0.5 of 87.8%. After that, XAI was implemented using the Grad-CAM and Eigen-CAM algorithms. Based on the experiments conducted, the Eigen-CAM algorithm is superior in terms of speed compared to the Grad-CAM algorithm. However, the Grad-CAM algorithm represents a better heatmap even though it requires longer computational time because it performs backpropagation to calculate the gradient of an object. |
|---|