metadata

library_name: YOLO
license: mit
tags:
  - YOLO
  - object-detection
  - dla
  - historical
metrics:
  - P
  - R
  - mAP@50
  - mAP@50-95
pipeline_tag: object-detection
version:
  - YOLOv8

YOLOv8 - PELLET Casimir Marius - Line + Illustration detection

This model predicts the following elements from document images:

text line
illustration

Model description

The model has been trained using the YOLOv8 library on PELLET Casimir Marius dataset. It has been trained on images with their dimensions equal to 640 pixels, starting from the YOLOv8m checkpoint.

Evaluation results

The model achieves the following results:

Set	Class	Images	Instances	P	R	mAP@50	mAP@50-95
train	all	80	881	0.707	0.823	0.825	0.457
	text line	80	843	0.801	0.73	0.781	0.275
	illustration	80	38	0.613	0.916	0.87	0.64
val	all	10	129	0.608	0.872	0.847	0.493
	text line	10	125	0.791	0.744	0.7	0.256
	illustration	10	4	0.424	1	0.995	0.731
test	all	10	125	0.767	0.881	0.932	0.556
	text line	10	122	0.903	0.763	0.868	0.331
	illustration	10	3	0.631	1	0.995	0.78

How to use?

Download the weights of this model
Refer to the Ultralytics documentation to use this model.