Browse categories
Explore
Fiverr Pro
English
$
USD
I will build custom yolov8 object tracking and edge ai vision systems
Eyad Arshad
About this gig
Need custom object tracking or real-time edge AI deployment? I train custom YOLOv8 models and build high-performance video analytics pipelines.
WHAT I DO:
Custom YOLOv8/v11 model training: object detection, segmentation, pose estimation.
Multi-object tracking integration using DeepSORT or ByteTrack for automated counting.
Model optimization via FP16 quantization, ONNX Runtime, or TensorRT.
Edge deployment configuration for NVIDIA Jetson, Raspberry Pi, or embedded Linux.
Live RTSP stream processing pipelines with event logging and database writes.
I do not just deliver a .pt file. I deliver compiled, hardware-optimized inference systems ready for edge deployment.
Basic: Custom model training on your dataset with a metrics report.
Standard: Tracking/counting pipeline for video feeds or RTSP streams.
Premium: Optimized edge-ready vision system (Jetson/Pi) with source code.
Message me with your hardware and dataset requirements.
Get to know Eyad Arshad
Eyad Arshad
AI Engineer specializing in Computer Vision, ML Systems, Python and CPP
- FromPakistan
- Member sinceOct 2018
- Avg. response time1 hour
Languages
Urdu, English
I am an AI engineer and C++ developer building production-ready desktop applications and computer vision systems. I don't build raw notebooks; I deliver optimized software engineered to survive real-world loads.
PROVEN WORK:
• HELIX: ML malware detection client with x86 instruction emulation (99.6% accuracy).
• Smart Traffic: Real-time YOLOv8 over ONNX on 4 camera feeds with Arduino hooks.
• UtiliSOFT: 7-module native C++/Qt6 enterprise desktop ERP layout.
Need a zero-lag PyQt6 app or clean native system? Message me before ordering—let's scope your architecture.
My Portfolio
FAQ
What hardware is this optimized for?
I optimize models for NVIDIA Jetson (Nano, Orin), Raspberry Pi 4/5, x86/x64 edge PCs, and embedded Linux devices.
How do you optimize models for edge devices with low compute?
I convert models to ONNX Runtime or TensorRT and apply quantization (FP16/INT8) to reduce model size and latency.
Can you handle multi-camera RTSP video feeds?
Yes. I build custom multi-threaded Python ingestion pipelines to pull, process, and track objects from multiple IP cameras simultaneously.
What object tracking algorithms do you support?
I integrate DeepSORT, ByteTrack, or BoT-SORT depending on your tracking accuracy and hardware constraints.
Do I need to provide annotated images for training?
If you have annotated data, that is best. If not, I can handle the annotation using CVAT or Roboflow (message me for a custom quote).
Can you turn the model into a desktop app?
Yes. Check my PyQt6 gig for professional desktop deployment, or message me for a combined offer.
