yolo8m-field-keypoints

Model Description

This is a YOLOv8-Medium model trained for field keypoint detection for football analytics.

Architecture: YOLOv8-Medium
Task: keypoint-detection
Image Size: 640
Classes: 1
Training Epochs: 20

Model Details

• Model Type: Object Detection
• Framework: PyTorch/Ultralytics
• Model Size: Medium
• Training Dataset: Football Field Keypoints Dataset

Classes

Detects key field markers and boundaries for spatial analysis

Usage

from ultralytics import YOLO
import requests
from PIL import Image

# Load the model
model = YOLO('hf://football-analytics/yolo8m-field-keypoints')

# Run inference
results = model('path/to/image.jpg')

# Process results
for result in results:
    boxes = result.boxes  # Bounding boxes
    if boxes is not None:
        for box in boxes:
            print(f"Class: {box.cls}, Confidence: {box.conf}")

Integration with Football Analytics Platform

This model is part of a comprehensive football analytics platform that provides:

• Player detection and tracking
• Field keypoint detection
• Advanced video analysis
• Real-time performance metrics

Training Configuration

agnostic_nms: false
amp: true
augment: false
batch: 8
box: 4.08816
boxes: true
cache: false
cfg: null
classes: null
close_mosaic: 10
cls: 0.29075
conf: null
copy_paste: 0.0
cos_lr: false
data: config_pitch_dataset.yaml
degrees: 0.0
deterministic: true
device: null
dfl: 1.32755
dnn: false
dropout: 0.0
dynamic: false
epochs: 20
exist_ok: false
fliplr: 0.36812
flipud: 0.0
format: torchscript
fraction: 1.0
freeze: null
half: false
hsv_h: 0.00554
hsv_s: 0.86584
hsv_v: 0.44793
imgsz: 640
int8: false
iou: 0.7
keras: false
kobj: 1.0
label_smoothing: 0.0
line_width: null
lr0: 0.00345
lrf: 0.00965
mask_ratio: 4
max_det: 300
mixup: 0.0
mode: train
model: yolov8m.pt
momentum: 0.84208
mosaic: 0.50009
name: train208
nbs: 64
nms: false
opset: null
optimize: false
optimizer: auto
overlap_mask: true
patience: 50
perspective: 0.0
plots: true
pose: 12.0
pretrained: true
profile: false
project: null
rect: false
resume: false
retina_masks: false
save: true
save_conf: false
save_crop: false
save_dir: runs\detect\train208
save_hybrid: false
save_json: false
save_period: -1
save_txt: false
scale: 0.08018
seed: 0
shear: 0.0
show: false
show_conf: true
show_labels: true
simplify: false
single_cls: false
source: null
split: val
stream_buffer: false
task: detect
tracker: botsort.yaml
translate: 0.02335
val: true
verbose: true
vid_stride: 1
visualize: false
warmup_bias_lr: 0.1
warmup_epochs: 4.32184
warmup_momentum: 0.87136
weight_decay: 0.00017
workers: 8
workspace: 4

Performance

Training metrics and validation results are included in the model files.

License

This model is released under AGPL-3.0 license.

Citation

If you use this model, please cite:

@misc{football-analytics-yolo,
  title={Football Analytics YOLO Model},
  author={Football Analytics Team},
  year={2024},
  url={https://huggingface.co/football-analytics/yolo8m-field-keypoints}
}