I will build yolo object detection and image classification models
AI and Machine Learning Engineer, Deep Learning, Computer Vision, NLP
About this Gig
Need to detect or classify objects in your images or videos? I train custom YOLO models on your own data and deliver a detector that is measured, explained and ready to run.
What I build
Object detection with YOLO11 and YOLOv8
Image classification models
Rare or safety critical classes, such as drowning or no helmet
Examples: pool safety, traffic rules, facial emotion
What you get
Trained weights and an inference script for images or video files
mAP50, precision and recall per class, confusion matrix and FPS
Your test images with predicted boxes and confidence scores
ONNX export in the top package
A plain English report
Proven results
0.874 mAP50, pool safety detection with 5 swimmer states including drowning
0.814 mAP50 at 33.70 FPS, helmet and vehicle detection
0.857 mAP50 at 55.63 FPS, facial emotion detection
Why choose me
Fast models that keep accuracy on rare classes
Honest numbers, always measured on images the model never saw
Free feasibility check before you order
Updates at every step and help running the code after delivery
Send me sample images and the objects you need to detect. I will reply with a plan.
Programming language:
Python
•
Colab
Tools:
Jupyter Notebook
•
OpenCV
•
Colab
Frameworks:
Scikit-learn
•
PyTorch
•
Panda
Other Data Science & ML Services I Offer
FAQ
Do I need labelled images?
Yes, YOLO needs bounding box labels. If you have Roboflow or YOLO format labels, send them. If not, message me first.
Which YOLO version will you use?
YOLO11 or YOLOv8. In the top package I train both and keep the better one for your data.
How do you handle rare classes?
I augment rare classes, for example 3x copies with flips, rotation and noise, to lift their recall.
Can it run on video files?
Yes. The inference script accepts images or video files, and I report the speed in FPS on test data.
Do I get the trained model?
Yes. You get the weights, the inference script and, in the top package, an ONNX export.

