yolov8-vs-yolo11

Evaluation Results

YOLOv8 vs YOLO11 AP

Model Size	YOLOv8 (mAP50-95)	YOLO11 (mAP50-95)	mAP50-95 Improvement (YOLO11 - YOLOv8)
N	0.371	0.392	0.021
S	0.447	0.467	0.020
M	0.501	0.514	0.013
L	0.529	0.532	0.003
X	0.540	0.547	0.007

---

YOLOv8 vs YOLO11 Parameters

Model Size	YOLOv8 Parameters (M)	YOLO11 Parameters (M)	Reduction Rate (%)
n	3.2	2.6	18.75%
s	11.2	9.4	16.07%
m	25.9	20.1	22.39%
l	43.7	25.3	42.09%
x	68.2	56.9	16.55%

---

YOLOv8 vs YOLO11 FLOPs

Model Size	YOLOv8 FLOPs (B)	YOLO11 FLOPs (B)	Reduction Rate (%)
n	8.7	6.5	25.29%
s	28.6	21.5	24.83%
m	78.9	68.0	13.81%
l	165.2	86.9	47.40%
x	257.8	194.9	24.40%

I didn't measure inference time of models because I was too lazy, and Ultralytics had already done it.

To examine the AP per class in detail, refer to the CSV files below:

---

• YOLOv8n.csv
• YOLOv8s.csv
• YOLOv8m.csv
• YOLOv8l.csv
• YOLOv8x.csv

---

• YOLO11n.csv
• YOLO11s.csv
• YOLO11m.csv
• YOLO11l.csv
• YOLO11x.csv

YOLOv8n vs YOLO11n

YOLOv8s vs YOLO11s

YOLOv8m vs YOLO11m

YOLOv8l vs YOLO11l

YOLOv8x vs YOLO11x

Fun Facts

ChatGPT4 played a significant role in helping me with this. I provided the prompts, and it handled the details.