Modelpublic

Arshia82sbn/pcb-yolov8s-detection

sourceHugging Faceupdated 2mo agoView on Hugging Face

0likes63downloads

Model Card

PCB Object Detection - YOLOv8s

A fine-tuned YOLOv8s model for detecting 21 types of PCB (Printed Circuit Board) components. Trained for 90 epochs on a custom PCB dataset with 1,788 training images and 173 validation images.

Model Details

Property	Value
Model	YOLOv8s (small)
Task	Object Detection
Input Size	640x640
Classes	21 PCB component types
Training Epochs	90
Optimizer	AdamW
Learning Rate	0.01
Batch Size	8
Seed	42
AMP	Enabled
Device	CUDA (GPU 0)

Performance Metrics (Epoch 69 - Best)

Metric	Value
Precision	0.761
Recall	0.571
mAP@0.5	0.619
mAP@0.5:0.95	0.400

Per-Class mAP@0.5

Class	mAP@0.5
relay	0.984
buzzer	0.953
display	0.933
ic	0.801
battery	0.799
switch	0.788
button	0.860
potentiometer	0.700
resistor	0.635
transistor	0.629
connector	0.602
capacitor	0.587
diode	0.556
clock	0.561
led	0.523
inductor	0.384
fuse	0.339
pads	0.012
pins	0.119
heatsink	0.000

21 PCB Classes

battery, button, buzzer, capacitor, clock, connector, diode, display, fuse, heatsink, ic, inductor, led, pads, pins, potentiometer, relay, resistor, switch, transformer, transistor

Training Configuration

yaml

task: detect
model: yolov8s.pt
epochs: 90
imgsz: 640
batch: 8
lr0: 0.01
lrf: 0.01
momentum: 0.937
weight_decay: 0.0005
warmup_epochs: 3.0
optimizer: auto
seed: 42
deterministic: true
amp: true
augmentations:
  hsv_h: 0.015
  hsv_s: 0.7
  hsv_v: 0.4
  fliplr: 0.5
  mosaic: 1.0
  erasing: 0.4

Usage

Installation

bash

pip install ultralytics

Inference

python

from ultralytics import YOLO

model = YOLO("best.pt")
results = model.predict("path/to/pcb_image.jpg", conf=0.25)

for result in results:
    boxes = result.boxes
    print(f"Detected {len(boxes)} objects")
    for box in boxes:
        cls = int(box.cls[0])
        conf = float(box.conf[0])
        xyxy = box.xyxy[0].tolist()
        print(f"  Class: {model.names[cls]}, Confidence: {conf:.3f}, Box: {xyxy}")

Export to ONNX

python

from ultralytics import YOLO

model = YOLO("best.pt")
model.export(format="onnx", imgsz=640, simplify=True)

Files

File	Description
`best.pt`	Best model checkpoint (epoch 69)
`last.pt`	Last epoch checkpoint
`best.onnx`	ONNX exported model
`results.csv`	Training metrics per epoch
`training_args.yaml`	Full training configuration
`results.png`	Training curves
`confusion_matrix.png`	Confusion matrix
`confusion_matrix_normalized.png`	Normalized confusion matrix
`BoxF1_curve.png`	F1-Confidence curve
`BoxP_curve.png`	Precision-Confidence curve
`BoxPR_curve.png`	Precision-Recall curve
`BoxR_curve.png`	Recall-Confidence curve
`labels.jpg`	Label distribution

Training Results

[image]

Dataset

This model was trained on the Detecting the PCB Object dataset from Roboflow Universe.

—Training images: 1,788
—Validation images: 173
—Test images: Available
—Image size: 640x640

Citation

If you use this model, please cite:

@misc{pcb_object_detection_yolov8,
  title={PCB Object Detection using YOLOv8s},
  author={Arshia},
  year={2026},
  publisher={Hugging Face},
  note={Fine-tuned YOLOv8s model for 21-class PCB component detection}
}

License

This model is released under the MIT License.