Parcourir les catégories
Explorer
Fiverr Pro
Français
$
USD
I will build custom object detection and computer vision models with yolo
Senior AI Developer
Machine Learning / AI Engineer with 3 years building production computer vision and LLM systems end to end. I train custom YOLO detection and segmentation models — including the hard small-object case...
À propos de ce service
Need a computer vision model that actually works on your real images not a toy demo?
I'm a machine learning engineer with 3 years building and shipping production computer vision systems end to end. I train custom object detection, instance segmentation, classification, pose and tracking models with YOLOv8/YOLO26, OpenCV and PyTorch and I specialise in the hard cases: tiny objects, dense scenes and high-resolution imagery where off-the-shelf models fail.
What I can build for you
Custom object detection & counting on your own dataset
Instance and semantic segmentation, pose/landmark estimation
Small-object detection via high-resolution tiling + IoU fusion
Object tracking, OCR, image classification
Clean Python, training scripts, a ReadMe and a FastAPI inference endpoint
Why me: I've raised validation mAP@50 from 0.73 to 0.93 on hard small-object data using tiling, built measurement and geometry pipelines, and deployed models in Docker on AWS. You get documented, reproducible, deployable work.
Message me before ordering so I can confirm scope, dataset and delivery.
Langage de programmation:
Python
•
Colab
•
MLflow
•
Amazon SageMaker
Outils:
Jupyter Notebook
•
opencv
•
MLflow
•
CVAT
•
PyTorch
Frameworks:
Scikit-learn
•
DeepPy
•
keras
•
PyTorch
•
Panda
Mon portfolio
FAQ
Do you work with my own images and dataset?
Yes — you provide the images. I can also handle labeling as an extra, or build a proof of concept on a public dataset first.
Which frameworks do you use?
YOLOv8/YOLO26, PyTorch, OpenCV and Ultralytics. TensorFlow or Detectron2 on request.
Can you detect very small objects?
Yes. High-resolution tiling with IoU fusion is my specialty — it's how I lift accuracy on objects that standard downscaled training misses.
Do I get the trained weights and source code?
Every package includes the model weights, training and inference scripts, and a ReadMe.
Can you deploy the model for me?
Premium includes a FastAPI endpoint and Docker setup. Deployment to your own AWS account is available as an extra.
Can you guarantee a specific accuracy?
No — accuracy depends on your data. I report honest metrics and push performance as far as the dataset allows, and I'll tell you early if the data won't support the goal.
