🌟 YoloV8s Dataset v0.2

Metadata

• Type: Yolov8s
• Filter: [64,128,256,512,768]
• Kernel Size: 3
• Train Size: 1400 × 1400
• Images Size: 963 × 1400

Preprocessing Images

• Auto-Orient: Applied
• Resize: Fit (white edges) in 963x1400
• Auto-Adjust Contrast: Using Contrast Stretching

Augmentations for Images

• Flip: Horizontal, Vertical
• Rotation: Between -15° and +15°
• Grayscale: Apply to 100% of images
• Exposure: Between -10% and +10%
• Blur: Up to 1px
• Noise: Up to 0.5% of pixels

Training data

• 480 Images (with generated images, original size 200)
• 420 for training (image set)
• 40 for validation (image set)
• 20 for test (image set)
• 4430 annotations (1846 without the generated images)

Train Accuracy

Precision(M)	Recall(M)	mAP50(M)	mAP50-95(M)	CLS Loss	DFL Loss
0.96524	0.97068	0.98096	0.91488	0.31534	1.04077