Browse categories
Explore
Fiverr Pro
English
$
USD
I will build computer vision, detection, analysis systems
Level 1
Has met certain performance criteria and shows strong potential in the marketplace.
About this Gig
Need software that can detect, track or understand what appears in images or video?
I build computer-vision pipelines around a clear business task and test them on your real examples. Depending on the problem, that may use object detection, segmentation, pose estimation, OpenCV processing or a vision-language model that turns visual input into structured JSON.
Recent work includes a Sports-video analyzer with 33-point pose tracking, batter lock-on and contact-frame detection. I also built a geospatial VLM pipeline that reviews street imagery, records asset condition and exports CSV, GeoJSON and an interactive map.
You can order help with:
- object detection and tracking
- image segmentation or classification
- pose estimation and sports video analysis
- OCR and visual data extraction
- VLM pipelines for images, documents or inspections
- evaluation, API delivery and deployment
You receive runnable code, setup notes and results measured on an agreed test sample.
Send a short description, I will tell you which package fits and what can be measured.
My Portfolio
Other Data Science & ML Services I Offer
FAQ
Can you train YOLO on my dataset?
Yes, when YOLO fits the task. I first check label quality, class balance and the target device. Training compute and annotation work are quoted separately when they exceed the selected package.
Do you build vision-language model workflows?
Yes. I can use a VLM to describe, classify or extract structured fields from images and documents. I add schema checks and confidence handling so uncertain cases can be reviewed.
Can you analyze video in real time?
Possibly. Share the target hardware, resolution, frame rate and latency goal. Real-time performance depends on the model, stream and device, so I benchmark before promising it.
Will you guarantee an accuracy score?
I will agree on a test set and report the measured result. A useful target depends on data quality, class definitions and how costly each error is.
Do I need labeled data?
Not for every job. A VLM or existing model may work without custom training. If labels are needed, I can define the annotation format and estimate the effort.
