Skip to content

Commit 03c0b93

Browse files
strng-ferstrng-fer
committed
Update .gitignore and enhance README.md with detailed project documentation
1 parent ad12f81 commit 03c0b93

File tree

3 files changed

+787
-12
lines changed

3 files changed

+787
-12
lines changed

‎.gitignore‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -97,4 +97,4 @@ runs/
9797
# Markdown files (except README.md)
9898
*.md
9999
!README.md
100-
!Step-by-Step_Guide.md
100+
!STEP_BY_STEP.md

‎README.md‎

Lines changed: 259 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -1,23 +1,271 @@
1-
# 🚧 MLOps Pothole Detection - YOLOv8
1+
# MLOps Pothole Detection System
2+
## Implementasi YOLOv8 dengan End-to-End MLOps Pipeline
23

3-
End-to-end MLOps pipeline for pothole detection using YOLOv8n and YOLOv8s with complete CI/CD, experiment tracking, and monitoring.
4+
---
5+
6+
## 👥 Tim Pengembang
47

5-
# Team
68
- Member 1: Feryadi Yulius (122450087)
79
- Member 2: Syadza Puspadari Azhar (122450072)
8-
- Member 3: Dinda Nababan 1224500
10+
- Member 3: Dinda Nababan (122450120)
911
- Member 4: Alyya 1224500
12+
---
13+
14+
## 📋 Latar Belakang
15+
16+
### Permasalahan
17+
- Kerusakan jalan (pothole) menimbulkan risiko kecelakaan
18+
- Deteksi manual tidak efisien dan memakan waktu
19+
- Perlu sistem otomatis untuk deteksi real-time
20+
21+
### Solusi
22+
- Implementasi Deep Learning dengan YOLOv8 untuk deteksi pothole
23+
- Pipeline MLOps untuk automasi training hingga deployment
24+
- Monitoring dan tracking untuk menjaga kualitas model
25+
26+
---
27+
28+
## 🎯 Tujuan Proyek
29+
30+
### Objektif Utama
31+
1. **Data Pipeline** → Automasi pengolahan data
32+
2. **Model Training** → Training dengan experiment tracking
33+
3. **Model Evaluation** → Validasi performa model
34+
4. **Deployment** → REST API & Container Docker
35+
5. **Monitoring** → Tracking performa production
36+
6. **CI/CD** → Automasi testing dan deployment
37+
38+
---
39+
40+
## 🏗️ Arsitektur Sistem
41+
42+
### Komponen Utama
43+
```
44+
┌─────────────┐ ┌──────────────┐ ┌─────────────┐
45+
│ Dataset │ --> │ Training │ --> │ MLflow │
46+
│ Validation │ │ Pipeline │ │ Tracking │
47+
└─────────────┘ └──────────────┘ └─────────────┘
48+
│
49+
▼
50+
┌─────────────┐ ┌──────────────┐ ┌─────────────┐
51+
│ User UI │ <-- │ FastAPI │ <-- │ Model │
52+
│ (Streamlit)│ │ Server │ │ Registry │
53+
└─────────────┘ └──────────────┘ └─────────────┘
54+
│
55+
▼
56+
┌──────────────┐
57+
│ Monitoring │
58+
│ System │
59+
└──────────────┘
60+
```
61+
62+
---
63+
64+
## 🔧 Teknologi yang Digunakan
65+
66+
### Machine Learning
67+
- **YOLOv8** → Model object detection
68+
- **Ultralytics** → Framework training
69+
- **PyTorch** → Deep learning engine
70+
71+
### MLOps Tools
72+
- **MLflow** → Experiment tracking & model registry
73+
- **Optuna** → Hyperparameter tuning
74+
- **DVC** → Data version control
75+
76+
### Backend & API
77+
- **FastAPI** → REST API server
78+
- **Streamlit** → User interface
79+
- **Docker** → Containerization
80+
81+
### CI/CD & Testing
82+
- **GitHub Actions** → Automation pipeline
83+
- **Pytest** → Unit testing
84+
- **Pre-commit** → Code quality
85+
86+
---
87+
88+
## 📊 Dataset
89+
90+
### Informasi Dataset
91+
- **Sumber**: Roboflow Pothole Detection Dataset
92+
- **Total Images**: 665 gambar
93+
- **Train**: 477 gambar (72%)
94+
- **Validation**: 143 gambar (21%)
95+
- **Test**: 45 gambar (7%)
96+
- **Classes**: 1 (Pothole)
97+
98+
### Preprocessing
99+
- Resize: 640x640
100+
- Normalization
101+
- Augmentation: flip, rotation, brightness
102+
103+
---
104+
105+
## 🚀 Pipeline MLOps
106+
107+
### 1. Data Ingestion
108+
- Validasi struktur dataset
109+
- Exploratory Data Analysis (EDA)
110+
- Quality checks
111+
112+
### 2. Training Pipeline
113+
- Multi-model training (YOLOv8n, YOLOv8s)
114+
- Hyperparameter tuning dengan Optuna
115+
- Experiment tracking dengan MLflow
116+
- Model versioning
117+
118+
### 3. Evaluation
119+
- **Metrics**: Precision, Recall, F1-Score, mAP
120+
- Confusion matrix
121+
- Test set evaluation
122+
- Model comparison
123+
124+
### 4. Deployment
125+
- Model serving via FastAPI
126+
- Docker containerization
127+
- Multi-model support
128+
- Load balancing
129+
130+
### 5. Monitoring
131+
- Inference latency tracking
132+
- Error rate monitoring
133+
- Input drift detection
134+
- Performance metrics
135+
136+
---
137+
138+
## 💻 Implementasi Teknis
139+
140+
### Model Training
141+
```python
142+
# CLI command
143+
python cli.py train --model yolov8n --epochs 100
144+
145+
# Hyperparameter tuning
146+
python cli.py tune --n-trials 50
147+
```
148+
149+
### Model Serving
150+
```python
151+
# Start API server
152+
python cli.py serve
153+
154+
# Access API: http://localhost:8000
155+
```
156+
157+
### Docker Deployment
158+
```bash
159+
docker-compose up --build
160+
161+
# Services:
162+
# - MLflow: http://localhost:5000
163+
# - FastAPI: http://localhost:8000
164+
# - User UI: http://localhost:8501
165+
# - Admin UI: http://localhost:8502
166+
```
167+
168+
---
169+
170+
171+
## 🎨 User Interface
172+
173+
### User App (Streamlit)
174+
- Upload gambar untuk deteksi
175+
- Real-time inference
176+
- Visualisasi hasil deteksi
177+
- Download hasil
178+
179+
### Admin App
180+
- Model management
181+
- Performance monitoring
182+
- Experiment comparison
183+
- System health check
184+
185+
---
186+
187+
## ✅ Testing & Quality Assurance
188+
189+
### Test Coverage
190+
- Unit tests: 85%
191+
- Integration tests
192+
- API endpoint tests
193+
- Data validation tests
194+
195+
### CI/CD Pipeline
196+
```
197+
Push → Tests → Build → Deploy
198+
↓ ↓ ↓ ↓
199+
Code Pytest Docker Production
200+
```
201+
202+
---
203+
204+
## 🔍 Monitoring & Observability
205+
206+
### Metrics Tracked
207+
1. **Model Performance**
208+
- Accuracy, Precision, Recall
209+
- Inference latency
210+
211+
2. **System Metrics**
212+
- API response time
213+
- Error rates
214+
- Resource usage
215+
216+
3. **Data Quality**
217+
- Input distribution
218+
- Drift detection
219+
220+
---
221+
222+
---
223+
224+
## 🎓 Lessons Learned
225+
226+
### Technical
227+
- Importance of experiment tracking
228+
- Docker containerization benefits
229+
- CI/CD automation value
230+
231+
### MLOps Best Practices
232+
- Version everything (code, data, model)
233+
- Monitor continuously
234+
- Automate repetitive tasks
235+
- Test thoroughly
236+
237+
---
238+
239+
## 🔮 Future Improvements
240+
241+
### Short Term
242+
- [ ] Model quantization untuk inference lebih cepat
243+
- [ ] Add more augmentation techniques
244+
- [ ] Improve UI/UX
245+
246+
### Long Term
247+
- [ ] Multi-class detection (berbagai jenis kerusakan)
248+
- [ ] Edge deployment (mobile/embedded)
249+
- [ ] Real-time video processing
250+
- [ ] Integration dengan GIS system
251+
252+
---
253+
254+
## 📚 Referensi
255+
256+
1. **YOLOv8 Documentation**: https://docs.ultralytics.com/
257+
2. **MLflow**: https://mlflow.org/
258+
3. **Base Project**: https://github.com/prsdm/mlops-project
259+
4. **FastAPI**: https://fastapi.tiangolo.com/
260+
261+
---
10262

11-
## 📚 Documentation
263+
## 📞 Kontak
12264

13-
**→ [START HERE: Step-by-Step Guide](STEP_BY_STEP.md)** ⭐
265+
**Repository**: https://github.com/sains-data/Mlops_RC_04
14266

15-
## 🎯 Project Objectives
16267

17-
- ✅ Build end-to-end MLOps pipeline for pothole detection
18-
- ✅ Implement CI/CD for Machine Learning
19-
- ✅ Experiment tracking and model monitoring
20-
- ✅ Provide User and Admin UI
268+
---
21269

22270
## 🏗️ Project Structure
23271

0 commit comments

Comments
 (0)