Browse categories
Explore
Fiverr Pro
English
$
USD
I will train custom yolo models for detection segmentation and pose estimation
About this Gig
I train custom YOLO models for detection, instance segmentation, pose/keypoints, classification and oriented bounding boxes (OBB). Each package covers ONE agreed task and dataset. Message me with sample images before ordering.
Includes label checks, fine-tuning, task-appropriate validation metrics, weights, training settings and Python inference code.
BASIC $100: 500 labeled images, 3 classes, 1 run, 7 days.
STANDARD $150: 1500 images, 5 classes, up to 3 runs, comparison and error review, 14 days.
PREMIUM $300: 3000 images, 10 classes, up to 5 runs, tested ONNX export and handoff guide, 21 days.
Labeling: $50/100 images (5 boxes/image, 5 classes), or $50/50 images (3 segmentation instances or pose objects/image, 3 classes). Each adds 3 days. Pose keypoints and mask complexity must be agreed first.
Training uses an NVIDIA RTX Pro 6000. Results depend on data quality; fixed accuracy is not guaranteed. Deployment, hosting and app development require separate scope.
Programming language:
Python
My Portfolio
FAQ
Do I need a labeled dataset?
Yes. Packages require labeled images. Extras: $50 for 100 images (max 5 boxes/image, 5 classes), or $50 for 50 images (max 3 segmentation instances or pose objects/image, 3 classes). Each adds 3 days. Agree keypoint count and mask complexity before ordering.
What will I receive?
Trained weights, training settings, environment requirements, validation metrics and a Python inference script. Standard adds run comparison and error review. Premium adds tested ONNX export and a handoff guide. App development, hosting and deployment will not be included.
Can you guarantee accuracy?
No fixed accuracy is guaranteed. I report task-appropriate metrics on an agreed validation split, such as detection/mask mAP, pose metrics or classification accuracy. Quality depends on data diversity and labels. Related frames and near-duplicates stay in the same split to avoid leakage.
What counts as a revision?
Revisions cover corrections within the agreed task, dataset and run allowance. Switching task type, adding datasets or classes, extra experiments, tracking or deployment requires additional scope. Errors in my delivered code will be corrected.
Which YOLO version and model size will you use?
We agree on version, model size and ONE task: detection, segmentation, pose, classification or OBB, based on sample images and target hardware. Keypoint schema and export compatibility are agreed before ordering. Framework and model licenses depend on the selected tools.
How were the portfolio examples produced?
PPE boxes and building masks are YOLO predictions. Yellow vehicle masks use SAM 2.1 Tiny with manually selected box prompts and review, not automatic YOLO vehicle detection. Poultry imagery illustrates pose/keypoint visualization. Dataset credits and workflow details are in my portfolio.
