YOLOv5: Training Results on the COCO128 Tutorial Dataset
In this article, we explore the YOLOv5 training results on the COCO128 tutorial dataset, and use Weights & Biases to keep track our experiments — effortlessly.
Created on November 2|Last edited on November 15
Comment
COCO128 is the tutorial dataset for YOLOv5, and it contains 128 images of COCO train 2017. In this article, we examine the YOLOv5 training results on the COCO128 tutorial dataset, and use Weights & Biases to track our experiments.
meta
["--batch","8","--weights","yolov5x.pt","--data","coco128.yaml","--epochs","300","--cache","--img","640","--nosave","--project","yolov5_tutorial","--name","COCO128-yolov5x"]
["--batch","16","--weights","yolov5l.pt","--data","coco128.yaml","--epochs","300","--cache","--img","640","--nosave","--project","yolov5_tutorial","--name","COCO128-yolov5l"]
["--batch","24","--weights","yolov5m.pt","--data","coco128.yaml","--epochs","300","--cache","--img","640","--nosave","--project","yolov5_tutorial","--name","COCO128-yolov5m"]
["--batch","32","--weights","yolov5s.pt","--data","coco128.yaml","--epochs","300","--cache","--img","640","--nosave","--project","yolov5_tutorial","--name","COCO128-yolov5s"]
{"remote":"https://github.com/ultralytics/yolov5","commit":"b3dabdcc380b45bbc802a3808457d8d0091e9148","__typename":"GitInfo"}
{"remote":"https://github.com/ultralytics/yolov5","commit":"41fdf9fa53bdc178f3df76764df5b655c94b6f7b","__typename":"GitInfo"}
{"remote":"https://github.com/ultralytics/yolov5","commit":"41fdf9fa53bdc178f3df76764df5b655c94b6f7b","__typename":"GitInfo"}
{"remote":"https://github.com/ultralytics/yolov5","commit":"41fdf9fa53bdc178f3df76764df5b655c94b6f7b","__typename":"GitInfo"}
1h 52m 19s
59m 11s
34m 56s
18m 20s
config
opt
8
16
24
32
COCO128-yolov5x
COCO128-yolov5l
COCO128-yolov5m
COCO128-yolov5s
yolov5_tutorial/COCO128-yolov5x
yolov5_tutorial/COCO128-yolov5l
yolov5_tutorial/COCO128-yolov5m
yolov5_tutorial/COCO128-yolov5s
yolov5x.pt
yolov5l.pt
yolov5m.pt
yolov5s.pt
8
16
24
32
COCO128-yolov5x
COCO128-yolov5l
COCO128-yolov5m
COCO128-yolov5s
yolov5_tutorial/COCO128-yolov5x
yolov5_tutorial/COCO128-yolov5l
yolov5_tutorial/COCO128-yolov5m
yolov5_tutorial/COCO128-yolov5s
yolov5x.pt
yolov5l.pt
yolov5m.pt
yolov5s.pt
summary
Bounding Box Debugger
metrics
0.97992
0.97758
0.97503
0.96707
0.90242
0.87807
0.8763
0.79979
train
Run set
4
System Reports Including GPU Utilization and GPU Temperature
Run set
4
Add a comment
snailfrying •
cool
Reply
Ayush Thakur • *
The side by side comparison of models is so informative. Thanks for the report Glenn.
Which model do you recommend to start with for a custom dataset? Is there a rule of thumb like start with a small model and then move to a larger model?
Reply
Glenn Jocher •
Very Cool!
1 reply
Iterate on AI agents and models faster. Try Weights & Biases today.