Browse categories
Explore
Fiverr Pro
English
$
USD
I will build a yolo object detection and tracking pipeline
Document AI and Workflow Automation Engineer
I build document extraction and n8n AI automation workflows that turn business inputs into structured outputs teams can review and use. My work includes invoice and PDF processing, API and spreadsheet...
About this Gig
Please message me before ordering. Send a few sample images or video clips and tell me what you want to detect or track.
I can check your data, labels, and target hardware. Depending on the package, I can build or fine-tune a YOLO detection model or a similar model, add tracking for one recorded video source, and provide Python code, model weights, test results, and setup notes.
I cannot promise a fixed accuracy or FPS before testing. If speed matters, I will report the model, image size, hardware, input, and test method used.
Please tell me which objects you need to detect, whether your data is labelled, what output you need, where the system should run, and how you will check the result.
Data collection, labelling, extra classes or cameras, API or app integration, cloud deployment, and new hardware are separate work unless we include them in the order.
Revisions cover problems within the agreed scope. After payment, you own my custom work. Pretrained models, datasets, and open-source tools keep their original licences.
My Portfolio
FAQ
I don't have a dataset yet. Can you help?
I can assess permitted data sources or frames from video you are authorized to use. Collection and labelling need a separate quote after checking source terms, privacy and licences. A sample review comes first; web availability alone does not establish permission.
Will this run on my Raspberry Pi or NVIDIA Jetson?
Send the exact device and sample input. I will check compatibility and feasibility before quoting. Edge deployment and optimization are custom work. Any speed report will state the hardware, model, image size and test method; no fixed FPS or lag-free result is guaranteed.
What does the Basic package include?
Basic tests one pretrained detector on one recorded clip after we confirm the classes and input. You receive annotated output, Python code, setup notes and limitations. Custom training, tracking and deployment are separate.
What files and usage rights do I receive?
You receive the agreed Python code, setup notes and test results. Trained weights are included when training is ordered. After full payment, you own my custom work. Pretrained weights, datasets and open-source tools remain subject to their licences.
Which models do you use?
I use a YOLO model suited to the input, required output and target environment. We agree the model and licence before work starts. Other architectures require a separate scope review. Model choice does not guarantee a particular accuracy or speed.
Can you integrate this into my app or website?
API integration, app integration, cloud hosting and edge deployment need a custom quote. We first agree the interface, target environment, access method, running costs and acceptance checks. These are not included in the standard packages.
What if I need to detect very small objects?
Send representative images or a short clip first. I will assess object size, image quality and label availability. Tiling or higher-resolution processing may help, but feasibility, compute cost and evaluation criteria must be agreed before ordering.
What are the training and video limits?
Standard: up to 500 labelled images, 3 classes and one run capped at 50 epochs or 4 GPU hours, whichever comes first, on buyer GPU. Premium adds tracking on one clip up to 5 minutes. Input is one camera, up to 1080p/30 FPS; that is an input limit, not processing speed.
What do revisions and support cover?
Revisions address issues within the agreed inputs, outputs and scope. New classes, footage types, cameras, platforms or features need a new quote. Package handoff includes the stated revision count; ongoing monitoring and maintenance are separate work.
What accuracy or tracking performance can I expect?
We agree a test sample and evaluation method before ordering. I report results and limitations, not a guaranteed score or FPS. Tracks may break during occlusion or missed detections. The output is for review; cross-camera identity matching and event tagging are separate work.
