File size: 2,816 Bytes
bd2737e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 |
---
license: gpl-3.0
datasets:
- Msun/modelnet40
language:
- en
metrics:
- accuracy
tags:
- deeplearning
---
# **Automated Defect Detection in 3D Mesh Files Using Multi-Model Deep Learning Approaches**
## **📌 Project Overview**
This project introduces a **multi-modal deep learning approach** to detect **defects in 3D mesh files** by combining:
- **CNN (Convolutional Neural Network)** for **object classification** using **ModelNet40** dataset images.
- **GNN (Graph Neural Network)** for **defect identification** in **OFF files (3D mesh models).**
- **Fusion Model** integrating **CNN and GNN** for improved classification accuracy.
## **📁 Dataset & Novelty**
The dataset used in this project is **novel and proprietary**, focusing on defect detection in 3D mesh files. Only the **ModelNet40 dataset** is publicly available.
### **🔹 Folder Structure**
```
📦 Dataset
┣ 📂 Images
┃ ┣ 📂 train
┃ ┃ ┣ 📂 category_1
┃ ┃ ┣ 📂 category_2
┃ ┃ ┗ ...
┃ ┗ 📂 test
┗ 📂 OFF_files
┣ 📂 train
┃ ┣ 📂 category_1
┃ ┃ ┣ 📂 normal
┃ ┃ ┗ 📂 defected
┃ ┣ 📂 category_2
┃ ┃ ┣ 📂 normal
┃ ┃ ┗ 📂 defected
┗ 📂 test
```
- **Images Folder** → Contains object images categorized into different classes (used for CNN).
- **OFF Files Folder** → Each category has **"normal"** and **"defected"** OFF files (used for GNN).
---
## **🚀 Model Architecture**
### **🔹 CNN Model (Image Classification)**
- Uses a **pretrained CNN model (ResNet)** to classify objects.
### **🔹 GNN Model (Defect Identification)**
- Processes **OFF files** using **node features** and **adjacency matrices**.
- Uses a **13-layer deep GNN model** to capture mesh structure defects.
### **🔹 Multi-Modal Fusion Model**
- Combines **CNN and GNN outputs** using **fully connected layers**.
- Improves **accuracy by leveraging both image and graph information**.
---
## **⚙️ Installation & Setup**
### **🔹 1️⃣ Install Dependencies**
```bash
pip install tensorflow numpy networkx trimesh
```
### **🔹 2️⃣ Run Training**
```bash
python Utils/train.py
```
### **🔹 3️⃣ Evaluate Model**
```bash
python Utils/evaluate.py
```
---
## **📊 Results & Evaluation**
- **CNN Classification Accuracy:** **76%**
- **GNN Defect Detection Accuracy:** **78%**
- **Fusion Model Accuracy:** **85%**
---
## **🛠️ Future Improvements**
- Use **a more complex GNN model** (with at least **13 layers**).
- Improve **multi-modal fusion model** by adding **extra layers**.
- Train on **a larger dataset** to improve generalization.
---
## **👨💻 Author**
**Dhanush**
📧 Contact: [e-mail](mailto:[email protected])
--- |