Download
1 variant available
This asset is only available as a PickleTensor which is a deprecated and insecure format. We caution against using this asset until it can be converted to the modern SafeTensor format.
320 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
(3)
Jul 19, 2026
Other
some update (train in imgsz 1280)
new classe organisation :
names:
0: gem_with_border_jewelry
1: loop_jewelry
2: bracelet
3: long_jewelry
4: ribbon_accessory
5: bell
6: large_jewelry
7: necklace
8: heart_emblem
9: cross_emblem
10: gem_jewelry
11: star_emblem
12: pearl_band
Show more

160 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
50 1 2 3 4 5 6 7 8 9
Jewelry detector (alpha) is a YOLO11s-seg model trained to detect and segment anime-style jewelry for ADetailer.
It can detect necklaces, long earrings/jewelry, bells, bracelets, gems, heart emblems, cross emblems, loop jewelry and some decorative accessories.
This is an early alpha release. It works on anime-style images, (i mean like anime screencap because i can't tell if semi realistic or stylized anime image will works as well), but small gems, thin chains, ribbons and ambiguous decorations may still be missed or misclassified.
Note: if the jewery is really poorly generated , it may not be detected.
Validation of 0.1:
34 images / 107 instances
Box:
P 0.730 / R 0.632 / mAP50 0.685 / mAP50-95 0.560
Mask:
P 0.730 / R 0.632 / mAP50 0.681 / mAP50-95 0.446
~190 training image
Classes:
- gem_jewelry
- loop_jewelry
- bracelet
- long_jewelry
- ribbon_accessory
- bell
- emblem_accessory
- necklace
- heart_emblem
- cross_emblem
may work better at imgsz = 1024 or 2048 in some case with the custom version of aadetailer (forge neo) in optional file.
use retina mask option for potential higher mask precision.

