UmaDiffusion/ULTIMA-YOLOv9

About ULTIMA-YOLO models

This is a part of ULTIMA project.

ULTIMA is Uma Musume Labeled Text-Image Multimodal Alignment.

ULTIMA-YOLOv9 model is a facial detection model for Uma Musumes in illustrations and based on yolov9-e and ULTIMA-YOLO dataset

This is the model repository for ULTIMA-YOLOv9, containing the following checkpoints:

• YOLO9-E

How to Use

Clone YOLOv9 repository.

git clone https://github.com/WongKinYiu/yolov9.git
cd yolov9

Download the weights using hf_hub_download and use the loading function in helpers of YOLOv9.

from huggingface_hub import hf_hub_download 
hf_hub_download("UmaDiffusion/ULTIMA-YOLOv9", filename="ultima_yolov9-e.pt", local_dir="./")

Load the model.

# make sure you have the following dependencies
import torch
import numpy as np
from models.common import DetectMultiBackend
from utils.general import non_max_suppression, scale_boxes
from utils.torch_utils import select_device, smart_inference_mode
from utils.augmentations import letterbox
import PIL.Image

@smart_inference_mode()
def predict(image_path, weights='ultima_yolov9-e.pt', imgsz=640, conf_thres=0.1, iou_thres=0.45):
    # Initialize
    device = select_device('0')
    model = DetectMultiBackend(weights=weights, device=device, fp16=False)
    stride, names, pt = model.stride, model.names, model.pt

    # Load image
    image = np.array(PIL.Image.open(image_path).convert("RGB"))
    img = letterbox(image, imgsz, stride=stride, auto=True)[0]
    img = img.transpose(2, 0, 1)
    img = np.ascontiguousarray(img)
    img = torch.from_numpy(img).to(device).float()
    img /= 255.0
    if img.ndimension() == 3:
        img = img.unsqueeze(0)

    # Inference
    pred = model(img, augment=False, visualize=False)

    # Apply NMS
    pred = non_max_suppression(pred[0][0], conf_thres, iou_thres, classes=None, max_det=1000)

or use detect.py in yolov9 repo.

python ./detect.py --source [image_path] --device 0 --img 1280 --weights './ultima_yolov9-e.pt' --name ultima_yolov9_1280_detect

Training Infomation

• Batch Size: 32
• Resolution: 640
• Epochs: 300, chosen best mAP
• GPU: 1x A6000 48GB
• Dataset: ULTIMA-YOLO

Statistics

• Train: 3,991 images
• Val: 399 images

Character Name	# in Train	# in Val	Precision	Recall	mAP50	mAP50-95
Agnes Tachyon	187	35	0.957	0.886	0.961	0.765
Air Groove	87	12	1	0.835	0.933	0.713
Air Shakur	75	12	0.986	1	0.995	0.909
Akikawa Yayoi	25	3	1	0.693	0.995	0.648
Admire Vega	74	16	1	0.754	0.894	0.707
Agnes Digital	50	6	0.992	0.833	0.972	0.803
Anshinzawa Sasami	25	1	0.956	1	0.995	0.796
Aston Machan	55	3	1	0.726	0.995	0.912
Bamboo Memory	41	3	0.97	1	0.995	0.895
Biko Pegasus	34	3	0.972	1	0.995	0.84
Byerley Turk	43	2	0.951	1	0.995	0.855
Bitter Glace	24	0	0.888	0.875	0.944	0.776
Biwa Hayahide	52	8	0.821	1	0.995	0.846
Copano Rickey	51	5	0.969	0.667	0.864	0.69
Curren Chan	54	9	0.996	1	0.995	0.801
Cheval Grand	43	13	0.998	1	0.995	0.783
Twin Turbo	120	13	0.982	1	0.995	0.842
Daiichi Ruby	57	5	0.963	1	0.995	0.949
Darley Arabian	48	2	1	0.837	0.995	0.819
Daring Tact	62	5	0.997	1	0.995	0.841
Daitaku Helios	100	11	1	0.903	0.961	0.787
Daiwa Scarlet	114	19	0.987	1	0.995	0.707
El Condor Pasa	65	6	0.951	1	0.995	0.808
Eishin Flash	39	5	0.853	1	0.995	0.927
Fuji Kiseki	48	6	1	0.875	0.995	0.88
Fine Motion	55	7	0.989	0.875	0.906	0.71
Gold City	49	8	0.942	0.938	0.991	0.81
Gold Ship	146	16	0.858	1	0.995	0.895
Godolphin Barb	44	2	0.84	0.833	0.851	0.659
Grass Wonder	74	6	1	0.797	0.995	0.792
Hishi Akebono	39	4	0.989	1	0.995	0.766
Hishi Amazon	46	6	0.993	1	0.995	0.835
Hayakawa Tazuna	34	5	1	0.659	0.922	0.638
Hishi Miracle	52	6	0.971	0.75	0.945	0.769
Happy Meek	51	4	1	0.787	0.938	0.808
Hokko Tarumae	50	9	1	0.678	0.995	0.76
Haru Urara	69	9	0.986	0.917	0.989	0.747
Ikuno Dictus	96	12	0.873	1	0.995	0.858
Ines Fujin	41	7	0.947	1	0.995	0.898
Inari One	46	2	0.856	1	0.995	0.656
Jungle Pocket	53	6	1	0.85	0.995	0.747
King Halo	77	6	0.975	1	0.995	0.773
Kashimoto Riko	34	3	1	0.778	0.995	0.823
Kiryuin Aoi	44	4	0.997	0.895	0.929	0.712
Kitasan Black	116	19	0.974	1	0.995	0.909
K.S.Miracle	48	3	0.982	1	0.995	0.852
Katsuragi Ace	43	4	0.989	1	0.995	0.881
Kawakami Princess	50	7	0.975	1	0.995	0.841
Little Cocon	51	3	1	0.567	0.995	0.796
Light Hello	25	2	0.993	1	0.995	0.788
Mr. C.B.	91	13	1	0.659	0.995	0.703
Meisho Doto	59	7	0.988	1	0.995	0.782
Mihono Bourbon	84	13	1	0.955	0.994	0.779
Manhattan Cafe	144	32	0.876	0.884	0.967	0.797
Mejiro Ardan	58	8	0.985	0.833	0.869	0.723
Mejiro Bright	55	6	0.987	1	0.995	0.813
Mejiro Dober	56	5	0.981	0.933	0.972	0.785
Mejiro McQueen	272	30	0.98	1	0.995	0.873
Mejiro Ryan	43	7	0.998	1	0.995	0.849
Matikanefukukitaru	52	7	1	0.952	0.995	0.719
Matikanetannhauser	87	13	0.996	1	0.995	0.81
Mejiro Palmer	95	11	0.893	1	0.929	0.822
Mejiro Ramonu	52	9	0.993	1	0.995	0.748
Maruzensky	43	7	0.984	1	0.995	0.684
Marvelous Sunday	40	6	1	0.702	0.995	0.668
Nakayama Festa	47	7	0.992	1	0.995	0.829
Nice Nature	96	8	0.993	1	0.995	0.723
Narita Brian	86	13	0.827	1	0.962	0.778
Narita Taishin	55	5	0.899	0.857	0.978	0.938
Nishino Flower	48	7	0.97	1	0.995	0.72
Narita Top Road	50	9	0.988	1	0.995	0.834
Oguri Cap	94	10	0.997	0.92	0.945	0.744
Rice Shower	165	25	0.992	1	0.995	0.89
Sakura Bakushin O	55	7	1	0.949	0.995	0.795
Symboli Rudolf	157	17	0.987	0.889	0.975	0.748
Sakura Chiyono O	48	9	0.946	0.8	0.941	0.835
Seiun Sky	72	10	0.98	1	0.995	0.842
Sakura Laurel	44	6	0.944	1	0.995	0.895
Shinko Windy	46	1	0.96	1	0.995	0.949
Seeking the Pearl	34	2	0.985	1	0.995	0.844
Symboli Kris S	68	6	0.87	0.958	0.943	0.728
Smart Falcon	53	7	0.976	1	0.995	0.876
Super Creek	48	4	1	0.959	0.995	0.736
Special Week	147	14	1	0.975	0.995	0.777
Silence Suzuka	129	18	0.993	1	0.995	0.84
Sirius Symboli	60	9	0.962	1	0.995	0.849
Satono Crown	47	2	0.993	0.75	0.925	0.746
Satono Diamond	79	12	0.98	0.75	0.775	0.649
Sweep Tosho	42	4	0.951	1	0.995	0.895
Tap Dance City	49	4	0.995	1	0.995	0.832
Taiki Shuttle	50	7	0.883	1	0.939	0.756
Tokai Teio	239	23	0.994	1	0.995	0.56
Tamamo Cross	59	6	1	0.86	0.99	0.748
T.M. Opera O	85	13	0.986	1	0.995	0.838
Tanino Gimlet	52	6	0.986	1	0.995	0.771
Mayano Top Gun	70	5	1	0.824	0.995	0.787
Tosen Jordan	68	9	0.959	1	0.995	0.801
Tsurumaru Tsuyoshi	38	2	0.984	1	0.995	0.736
Neo Universe	47	5	1	0.806	0.945	0.753
Vodka	110	15	0.954	1	0.995	0.895
Wonder Acute	53	1	0.976	0.8	0.962	0.877
Winning Ticket	47	5	0.997	1	0.995	0.889
Yukino Bijin	44	7	1	0.965	0.995	0.904
Yaeno Muteki	39	5	0.975	1	0.995	0.932
Yamanin Zephyr	42	3	0.976	0.714	0.96	0.747
Zenno Rob Roy	51	7	0.958	1	0.995	0.895
Furioso	15	0	0.938	1	0.995	0.995
Transcend	40	2	0.964	1	0.995	0.796
Espoir City	30	1	0.939	1	0.995	0.895
North Flight	40	2	0.946	1	0.995	0.597
Dantsu Flame	30	1	0.878	1	0.995	0.895
No Reason	26	0	0.961	0.667	0.699	0.53
Still in Love	28	1	0.961	1	0.995	0.895
Samson Big	25	1	0.891	1	0.995	0.697
Sounds of Earth	53	3	0.972	1	0.995	0.857
Royce and Royce	30	2	0.942	1	0.995	0.398
Duramente	43	1	0.939	1	0.995	0.895
Rhein Kraft	31	3	0.975	1	0.995	0.799
Cesario	37	1	0.947	1	0.995	0.796
Air Messiah	23	1	0.964	1	0.995	0.927
Daring Heart	28	0	0.961	1	0.995	0.858
Orfevre	25	3	0.947	1	0.995	0.995
Gentildonna	40	1	0.944	1	0.995	0.597
Win Variation	21	2	0.94	1	0.995	0.895
Venus Paques	37	2	0.935	1	0.995	0.796
Rigantona	28	1	0.995	1	0.995	0.91
Sonon Elfie	29	1	0.994	1	0.995	0.815