File size: 2,063 Bytes
1f1f275
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: mit
tags:
  - pytorch
  - onnx
  - object-detection
  - yolo
  - computer-vision
  - pcb-inspection
  - industrial
---

# Inspection Engine

## Model Description
Inspection Engine is a YOLO11s model fine-tuned for automated PCB (Printed Circuit Board) defect detection. It identifies manufacturing defects such as missing components, solder bridges, lifted pads, and other anomalies in real-time, enabling automated quality control in electronics manufacturing.

## Model Architecture
- **Base Model**: YOLO11s (Small variant — optimized for speed)
- **Framework**: PyTorch + ONNX (for cross-platform deployment)
- **Task**: Object Detection (Defect Localization)
- **Input**: High-resolution PCB images

## Training Details
- **Dataset**: PCB defect inspection dataset with annotated defect regions
- **Checkpoint**: `inspection_engine_final3` (best performing run)
- **Augmentations**: Mosaic, color jitter, random affine transforms
- **Export**: Exported to ONNX for production deployment

## Files
| File | Description |
|------|-------------|
| `best.pt` | Best PyTorch model weights |
| `best.onnx` | ONNX export for cross-platform/production deployment |

## Usage

### PyTorch (Ultralytics)
```python
from ultralytics import YOLO
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(repo_id='devanshty/Inspection-Engine', filename='best.pt')
model = YOLO(model_path)
results = model('pcb_image.jpg')
results[0].show()
```

### ONNX Runtime
```python
import onnxruntime as ort
import numpy as np
from huggingface_hub import hf_hub_download

onnx_path = hf_hub_download(repo_id='devanshty/Inspection-Engine', filename='best.onnx')
session = ort.InferenceSession(onnx_path)
# Prepare input (1, 3, H, W) float32 normalized
input_name = session.get_inputs()[0].name
outputs = session.run(None, {input_name: np.zeros((1, 3, 640, 640), dtype=np.float32)})
```

## Download & Use

```python
from huggingface_hub import hf_hub_download
model_path = hf_hub_download(repo_id='devanshty/Inspection-Engine', filename='best.pt')
```