Performance Comparison of YOLOv7 and YOLOv8 Using the YCB Datasets YCB-M and YCB-Video

EasyChair Preprint 13111, version 2

Abstract

In this paper, the two frameworks YOLOv7 and YOLOv8
are compared using two labeled YCB datasets YCB-M and YCB-Video.
We provide an additional dataset, called Robot Domain Dataset (RDD),
to evaluate the performance of the two YOLO frameworks on a new
data domain, to simulate situations were it is not possible to retrain
the models due to a lack of data or time. Furthermore, the impact of
different amounts of training data on performance is observed. For comparability,
the training and validation pipelines are provided. We were
able to show that both frameworks perform very well on the datasets we
retrained on. But on our new dataset YOLOv7 significantly outperforms
YOLOv8 by 22% mean average precision. The division of the datasets,
the code of the training and validation pipelines, the trained models and
the dataset RDD can be found at https://github.com/iki-wgt/yolov7_
yolov8_benchmark_on_ycb_dataset

Keyphrases: MS COCO, Service Robotics, YCB Dataset, YOLO, benchmark, map, object detection

@booklet{EasyChair:13111,
  author    = {Samuel Hafner and Markus Schneider and Benjamin Stähle},
  title     = {Performance Comparison of YOLOv7 and YOLOv8 Using the YCB Datasets YCB-M and YCB-Video},
  howpublished = {EasyChair Preprint 13111},
  year      = {EasyChair, 2024}}