Hugging Face's logo

Spaces:
Wicky
/
bus-inspection-classifier
Sleeping

App Files Community
bus-inspection-classifier
File size: 3,073 Bytes
60965e7
 
 
 
 
 
05becdd
60965e7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
05becdd
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
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
---
title: Bus Inspection Classifier
emoji: 🚌
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.2.0
app_file: app.py
pinned: false
license: mit
---

# 🚌 Bus Inspection Classifier - SigLIP v2 ONNX

An automated bus component classification system using the **SigLIP v2** vision model, optimized with ONNX Runtime for fast inference.

## 🎯 Model Details

- **Model**: SigLIP v2 (Sigmoid Loss for Language-Image Pre-training)
- **Format**: ONNX (372.6 MB)
- **Input**: RGB images (224x224)
- **Classes**: 18 bus component categories
- **Inference Time**: ~50-100ms per image (CPU)

## 📋 Supported Classes

The model can classify the following bus components:

1. AC Mat
2. Alco brake camera
3. Alco-brake device
4. Back windshield
5. Bus back side
6. Bus front side
7. Bus side
8. Cabin
9. Driver grooming
10. First aid kit
11. Floormats & POS
12. Front windshield
13. Hat rack
14. ITMS Device
15. Jack & Spare tyre
16. Luggage compartment
17. RFID Card
18. Seats

## 🔌 API Usage

### Using the Gradio Client (Recommended)

```python
from gradio_client import Client

# Connect to the Space
client = Client("YOUR-USERNAME/bus-inspection")

# Make a prediction
result = client.predict("path/to/bus_image.jpg")

print(result)
# Output: {
#   "class_name": "Bus front side",
#   "confidence": "98.45%",
#   "inference_time_ms": "87.32"
# }
```

### Using Python Requests

```python
import requests
import json
from PIL import Image
import io

# Your Space URL
API_URL = "https://YOUR-USERNAME-bus-inspection.hf.space/api/predict"

# Load image
with open("bus_image.jpg", "rb") as f:
    files = {"data": f}
    response = requests.post(API_URL, files=files)

result = response.json()
print(f"Class: {result['class_name']}")
print(f"Confidence: {result['confidence']}")
print(f"Inference Time: {result['inference_time_ms']} ms")
```

### Using cURL

```bash
curl -X POST https://YOUR-USERNAME-bus-inspection.hf.space/api/predict \
  -F "data=@bus_image.jpg"
```

## 📊 Response Format

The API returns a JSON object with the following fields:

```json
{
  "class_name": "Bus front side",
  "confidence": "98.45%",
  "inference_time_ms": "87.32"
}
```

- **class_name**: The predicted bus component category
- **confidence**: Model confidence score as a percentage
- **inference_time_ms**: Total inference time in milliseconds

## 🚀 Example Use Cases

- **Quality Control**: Automated inspection of bus components during manufacturing
- **Maintenance Checks**: Verify proper installation of safety equipment
- **Documentation**: Automatically categorize bus inspection photos
- **Training**: Help inspectors identify components correctly

## 🛠️ Technical Implementation

- **Framework**: Gradio for UI and API
- **Runtime**: ONNX Runtime (CPU optimized)
- **Preprocessing**: SigLIP normalization (mean=0.5, std=0.5)
- **Deployment**: Hugging Face Spaces with Git LFS

## 📝 License

MIT License - Free for commercial and personal use

## 🙏 Acknowledgments

Built with SigLIP (Sigmoid Loss for Language-Image Pre-training) from Google Research.