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')
```
|