I will build a custom ai safety detector for helmets and vests
Data Analyst in Python and Data Visualization
About this Gig
I'll build you a working custom object detection model using YOLOv8 trained specifically on your images, not a generic pre-built one.
I recently built a real-time PPE safety detection system (helmet/vest detection) from scratch, including dataset preparation, training, and tuning for real-world accuracy issues like lighting and distance. I bring that same hands-on process to your project whether you need to detect products, defects, people, vehicles, or any custom object class.
What you get:
- A YOLOv8 model trained on your provided dataset (or I'll help you source/label one if needed)
- Real-time inference setup so it runs on images or video
- A short report explaining model accuracy and where it performs well vs. needs more data
- Clean, documented code you can actually use or hand to your dev team
Why work with me: I'm early in my freelance career, which means your project gets my full attention, fast turnaround, and a rate well below what senior CV engineers charge without cutting corners on the actual training process.
Send me a sample of your images before ordering and I'll tell you honestly whether YOLOv8 is the right fit and roughly what accuracy is realistic no surprises after you
APIs:
Other
Programming language:
Python
•
SQL
Tools:
Jupyter Notebook
•
OpenCV
•
Excel
•
PyTorch
Frameworks:
Scikit-learn
•
PyTorch
•
Panda
My Portfolio
FAQ
What kind of objects can you detect?
Anything with enough labeled training examples like people, PPE/safety gear, vehicles, products, defects, etc. Send me a few sample images before ordering and I'll tell you honestly if YOLOv8 is a good fit.
Do I need to provide a labeled dataset?
Not necessarily. If you have raw images only, I can help with labeling, but it adds time and may cost more depending on dataset size. Let's discuss before you order.
What accuracy can I expect?
Depends heavily on your dataset size and quality. I'll give you a realistic accuracy estimate after reviewing your data, not before. I won't promise a number I can't actually hit.
Will this run in real-time on video?
Yes, the delivered model includes a real-time inference setup for video/webcam input.
What if the first results aren't good enough?
I'll share sample detection results with you during the project before final delivery, so we can discuss and adjust if something's off. you're not seeing the model for the first time only at the very end.

