Browse categories
Explore
Fiverr Pro
English
$
USD
I will train an object detection model and count objects in video
Will deliver computer vision, image processing, and opencv projects with python
CS student at North South University, Dhaka, working in computer vision. I train and deploy real-time detection models — YOLO, custom dataset annotation, OpenCV and MediaPipe pipelines.
Recent builds...
About this Gig
Most CV sellers show you a demo on a stock photo. I'll show you results on your footage before you order.
Send me 510 frames from your actual video. Within 24 hours I'll send back annotated detections and an honest verdict on whether this is solvable at the accuracy you need. No charge, no obligation. If it won't work, I'll tell you and save you the money.
What you get
- YOLOv8 model trained on your data weights are yours
- Dataset annotation and cleanup in Roboflow
- Metrics that mean something: mAP50-95, per-class precision/recall, confusion matrix
- A runnable inference script image, video, webcam or RTSP not a notebook you can't deploy
- Object counting, zone entry, and line-crossing alerts
- ONNX/TFLite export for Raspberry Pi, Jetson or mobile
- README so you can retrain it yourself when your data changes
Why me
I build the system around the model, not just the model. My last project was a dual-YOLOv8 engine doing live detection with a WebSocket dashboard, PostgreSQL logging and temporal alert filtering to kill false positives. Most detection gigs stop at the .pt file. Yours won't.
Message me with a few sample frames and what you need detected. I answer within the hour.
APIs:
Other
Programming language:
Python
Tools:
Jupyter Notebook
•
OpenCV
•
Colab
•
PyTorch
Frameworks:
Scikit-learn
•
PyTorch
•
Panda
•
Other
My Portfolio
FAQ
Do I really get a free test before ordering?
Yes. 5–10 frames, annotated and returned within 24 hours, plus my honest read on feasibility. It costs me an hour and saves us both a failed order.
You have no reviews. Why should I risk it?
You shouldn't have to. That's what the free test is for — you see real output on your own data before any money moves.
I have no labeled data. Can you still help?
Yes — annotation is included in Detection and Deployment. Send raw images or video and I'll label them.
What accuracy will I get?
Depends entirely on your data quality, class similarity, and lighting. I'll give you a realistic number after the free test, not before. Anyone quoting you 99% without seeing your footage is guessing 😅.
Will it run on a Raspberry Pi / Jetson / phone?
Usually yes, with a smaller model and ONNX or TFLite export. Tell me the exact hardware first and I'll tell you the realistic FPS.
Do I own everything?
Yes!! weights, scripts, annotated dataset, all yours. No licensing, no lock-in.
