Page 84 - Kỷ yếu hội thảo khoa học lần thứ 12 - Công nghệ thông tin và Ứng dụng trong các lĩnh vực (CITA 2023)

68

Fig. 2. YOLO V7 as compared to other YOLO versions [5]

The major models in the YOLO7 series are the YOLOv7-X, YOLOv7-E6,
YOLOv7-D6 and YOLOv7-E6E. All of these models beat the corresponding YOLOR
models with a similar or lesser number of parameters and gave APs of 52.9%, 55.9%,
56.3%, and 56.8%, respectively.

YOLO V8. YOLOV8 is the latest YOLO-based object show model from Ultralytics
provided with state-of-the-art performance. There are five models in each YOLOV 8
model categoryfor detection, segmentation, and classification. The YOLOV8 Nano is
the fastest and smallest, while the YOLOV8 Extra Large (YOLOV8-X) is the most
accurate but slowest of them: YOLOV8-n, YOLOV8-s, YOLOV8-m, YOLOV8-l,
YOLOV8-x.

Fig. 3. Compare YOLO V8 with YOLO V5, 6, 7 versions

In summary, from Figure 4, it is clear that the YOLO V8 series models seem to
perform much better than previous YOLO models, not only the YOLO V5 models but
also the YOLO V8 models are ahead of the YOLO V7 and YOLO V6 models.
The following will be the installation and experimental process to compare the
results when applying YOLO versions to the Hot Weapon identification problem.

CITA 2023 ISBN: 978-604-80-8083-9