
Closed
Posted
I need a complete computer-vision pipeline that lets me watch multiple indoor and outdoor cameras and immediately know who is where. The core goal is security surveillance, so every component has to run in real time without noticeable lag. Here is what the system must do: • Detect every person entering the frame, day or night, in varied lighting and weather. • Track each individual seamlessly across several overlapping or non-overlapping cameras (multi-object tracking). • Re-identify the same person when they reappear on a different camera, even after minutes out of view. • Expose a live stream API or dashboard that shows bounding boxes, unique IDs, and confidence scores as events unfold. I am open to the specific model stack—YOLO, Detectron2, DeepSORT, StrongSORT, or any other proven combination—as long as you can meet real-time FPS on standard GPU hardware and keep false positives low. Please factor in calibration for both indoor and outdoor lenses, simple configuration files for adding new cameras, and a straightforward way to export clips or snapshots if a tracked ID triggers an alert. Acceptance criteria: 1. End-to-end demo on at least two indoor and two outdoor cameras running simultaneously at 25 fps+. 2. Consistent ID hand-off across cameras with ≤5 % ID-switch rate on a one-minute test sequence. 3. Setup script and concise readme so I can replicate results on my own machine. If you’ve built something similar or have benchmark videos to prove speed and accuracy, that will move things forward quickly.
Project ID: 40711923
65 proposals
Remote project
Active 2 hours ago
Set your budget and timeframe
Get paid for your work
Outline your proposal
It's free to sign up and bid on jobs
65 freelancers are bidding on average $20 USD/hour for this job

I am a skilled computer vision specialist with extensive experience in developing real-time surveillance systems. I have implemented complex person tracking solutions using models like YOLO and DeepSORT, ensuring seamless multi-object tracking and minimal lag, essential for effective security monitoring. I have a robust understanding of your requirements for a multi-camera setup with real-time performance targets. Utilizing tools such as Detectron2 and StrongSORT, I can build a reliable pipeline that detects and re-identifies individuals accurately across various conditions, reducing ID-switch rates significantly. My experience with handling diverse lighting, weather, and hardware configurations allows me to tailor systems for both indoor and outdoor environments effectively. My recent projects include building a scalable video analytics platform that achieved 30 fps on standard GPUs with real-time streaming APIs, similar to your needs. I am keen to discuss how I can help meet your goals and answer any questions you might have. Could you please share more about the specific camera models and hardware you intend to use?
$20 USD in 40 days
8.4
8.4

Confident and Committed to Delivering Impeccable Results As an electrical engineer with years of hands-on experience, I am confident that I am the ideal candidate for your real-time CCTV person tracking project. My extensive experience in the aspects of embedded systems, firmware development, computer vision, and deep learning aligns perfectly with the core objectives you have set forth. Having completed numerous projects from ideation to prototype to final product, my skills extend seamlessly to every requirement on your list. From handling diverse microcontrollers, including those from TI and NXP, to firmware development in C and C++ and implementation of advanced algorithms for data analysis and IoT connectivity, I possess a holistic understanding of the entire product development workflow. Furthermore, my knowledge of PCB design and RF hardware will ensure that your system is accurately calibrated across your indoor and outdoor lenses. Importantly, I have a demonstrated ability to integrate AI/ML into edge devices efficiently – a crucial aspect for this project – therefore guaranteeing high-speed performance while keeping false positives at a minimum.
$25 USD in 40 days
8.3
8.3

⭐⭐⭐⭐⭐ Build a Real-Time Computer Vision Pipeline for Security Surveillance ❇️ Hi My Friend, I hope you're doing well. I've reviewed your project needs and see you're looking for a complete computer-vision pipeline. You don’t need to look any further; Zohaib is here to help you! My team has successfully completed 50+ similar projects in security surveillance. I will create a system that detects, tracks, and identifies individuals across multiple cameras in real time. I will utilize advanced models like YOLO or Detectron2 to ensure high performance. The system will support calibration for indoor and outdoor setups, and I will provide a user-friendly API for live streaming and alerts within the same budget. ➡️ Why Me? I can easily build your computer vision system as I have 5 years of experience in machine learning, computer vision, and real-time processing. My skills include object detection, multi-object tracking, and API development. Additionally, I have a strong grip on various frameworks like TensorFlow and OpenCV, ensuring a robust solution. ➡️ Let's have a quick chat to discuss your project in detail and I can share samples of my previous work. I look forward to discussing this with you! ➡️ Skills & Experience: ✅ Computer Vision ✅ Object Detection ✅ Multi-Object Tracking ✅ Real-Time Processing ✅ YOLO ✅ Detectron2 ✅ DeepSORT ✅ StrongSORT ✅ API Development ✅ Python Programming ✅ OpenCV ✅ Machine Learning
$17 USD in 40 days
7.9
7.9

Hi, I am a software engineer with over 16 years of experience, including real-time computer vision, GPU inference, video-stream processing, multi-object tracking, and production surveillance systems. I can build the complete pipeline for person detection, persistent tracking, cross-camera re-identification, and live display of IDs, boxes, and confidence scores across your indoor and outdoor feeds. I would begin by benchmarking a proven detector and tracker/re-ID stack against your actual camera conditions, then optimize inference, frame handling, and camera hand-off to sustain 25+ FPS across four concurrent streams on the target GPU. The solution will include lens/camera calibration, configuration-based camera onboarding, alert-triggered clip and snapshot export, a live API/dashboard, setup automation, and concise documentation. I will also validate FPS, false positives, and ID-switch rate against repeatable test sequences. Relevant projects and benchmark material cannot be shared publicly, but I can show suitable examples privately. Which GPU and camera/stream protocols will be used, and can you provide representative day/night footage plus a map of camera overlap and transition paths? Please contact me to discuss details.
$25 USD in 40 days
7.6
7.6

Hi, I understand you need a real-time multi-camera computer-vision pipeline for person detection, tracking and re-identification across indoor and outdoor cameras, with low latency and reliable ID continuity. I can build the pipeline around a GPU-optimised detector such as YOLO with a proven multi-object tracker and Re-ID model, selecting the final combination based on your camera resolution, GPU and accuracy requirements. I’ll separate detection, tracking and cross-camera Re-ID so each component can be benchmarked and tuned independently. The system will support simultaneous camera streams, configurable camera definitions and calibration parameters, persistent person IDs across overlapping and non-overlapping views, confidence scoring and live event output through an API or dashboard. I’ll also implement alert-triggered snapshot/clip capture and structured event logging. For validation, I’ll benchmark the complete pipeline across two indoor and two outdoor streams, measuring FPS, latency, detection quality and ID-switch rate. I’ll tune thresholds and Re-ID matching to target the specified ≤5% ID-switch rate while avoiding unrealistic performance claims where camera conditions or hardware impose limitations. I’ll provide the complete source, setup script, configuration examples and concise README so the environment can be reproduced and additional cameras added without restructuring the system. Best regards, Houssame
$17.50 USD in 39 days
6.8
6.8

I got you! I can build your real-time multi-camera surveillance pipeline with person detection, multi-object tracking, cross-camera re-identification, live API/dashboard overlays, camera config files, and clip/snapshot export for alerts. I’m ready to handle the full project specifics, including indoor/outdoor calibration, low-latency GPU inference, and testing against your 25 FPS+ and ≤5% ID-switch acceptance targets. I’m young, a fast learner, and available 24/7. Get the demo first before you pay. A couple of key questions: what GPU/server specs will the system run on, and do the four cameras have overlapping views or mostly separate entry/exit zones? Also, do you prefer a web dashboard, RTSP/HTTP stream output, or both? Let’s chat and discuss the answers so I can propose the best model stack and demo plan. Kind regards, Haroon Z
$25 USD in 1 day
6.0
6.0

As a team with a clear niche in agentic AI, we strongly feel that our experience of building end-to-end AI systems in real-world environments sets us apart from other candidates. By incorporating prolific technologies like Odoo ERP, IoT hardware design and manufacturing, and deploying on various cloud platforms, we are adept at seamlessly combining AI models with the existing workflows to make concrete decisions. We also possess a wealth of expertise in the full web and mobile stack which ensures that our systems remain edge-device compatible, ERP-integrated, and capable of handling live sensor data - an essential skillset for your project. Moreover, our capabilities extend beyond just coding and algorithmic implementations. Our solid foundation in hardware such as PCB design, ESP32/STM32 firmware, and MQTT-connected sensor networks is proof of our expertise in creating intelligent systems that can communicate effectively with diverse devices and capture real-time information.
$20 USD in 40 days
6.4
6.4

I’ll build you a real-time person tracking pipeline that detects, tracks, and re-identifies individuals across indoor and outdoor cameras with low latency and minimal ID switches. I’ll handle camera calibration, day/night robustness, and false-positive reduction through a proven YOLO + StrongSORT setup, and wrap it in a simple live dashboard with config-driven camera additions and alert-triggered clip export. The result will be a replicable, documented system that meets your 25 fps and ID-handoff requirements on standard GPU hardware.
$20 USD in 40 days
6.2
6.2

I will build the pipeline around multi-camera detection, tracking and re-identification, with GPU-optimised inference, camera calibration, configurable camera profiles, confidence scoring and ID handoff across overlapping and non-overlapping views. I will also expose the live tracking data through an API/dashboard and support alert-triggered snapshots or clips. I will benchmark the system against your 25+ FPS and ID-switch targets rather than relying on model accuracy alone. One important question: what GPU hardware will the final system run on? This will determine the model size, inference strategy and achievable FPS across four simultaneous cameras. Please ping me to get started and get outstanding results. Thanks!!!
$20 USD in 40 days
5.8
5.8

You need a computer vision system that tracks individuals across multiple indoor and outdoor camera feeds without losing their identity when they exit and re-enter different frames. The main difficulty is maintaining a consistent ID at 25 FPS across non-overlapping views while lighting conditions shift, which I will address by implementing a robust feature-matching Re-ID step alongside the detection pipeline. I can build this using a Python-based workflow, utilizing efficient models for detection and tracking to ensure low latency on your GPU hardware. I will provide a configuration-driven setup that allows you to add new cameras easily and export snapshots when specific IDs are detected. To ensure the 5% ID-switch target is met, I can deliver a small test run on two of your camera feeds first to verify the accuracy before we finalize the dashboard and alert system. Are your cameras currently providing an RTSP stream for the integration?
$20 USD in 14 days
4.5
4.5

Hi there, As a seasoned professional, there's not much I haven't seen and conquered when it comes to handling complex AI and software challenges. Combing through my extensive repertoire of skills, I noticed that this particular role aligns snugly with my experience in C programming, Deep Learning, Image Processing and Python, as well as my expertise in deploying the YOLO model stack. During my 55 projects executed with precision and speed, I've had the opportunity to develop a deep understanding of real-time systems that can handle complex image processing assignments like yours. My focused and streamlined approach ensures that your core goal of security surveillance is met, down to the last minute detail. With me, you can relax knowing that I will keep false positives low while still hitting those crucial high FPS benchmarks. Best regards, Mobasher Reza
$20 USD in 40 days
4.2
4.2

Tracking a person across multiple camera feeds, indoor and outdoor, is where most builds fall apart at the handoff between cameras. I would run YOLO for detection with a re-identification model so the person stays tagged as they move between feeds, built in Python and OpenCV. I can start today, working version in 4 days. Pricing and timeline here are starting points based on the post, we will lock in real numbers after a quick scope call.
$25 USD in 14 days
3.7
3.7

Hi there, I can develop a comprehensive computer vision pipeline that meets your requirements for real-time person tracking across multiple CCTV cameras. By utilizing proven models like YOLO or DeepSORT, I will ensure reliable detection, tracking, and re-identification of individuals across overlapping camera views, achieving the necessary FPS and maintaining a low ID-switch rate. I will provide a complete setup script and a concise readme for easy replication on your machine, alongside an end-to-end demo of the system using multiple cameras. Your satisfaction is my top priority, and I guarantee high-quality results. Regards, Ali
$20 USD in 1 day
3.7
3.7

Hi, there. "REAL-TIME CCTV PERSON TRACKING" — you need a complete computer vision surveillance pipeline that can detect, track, and re-identify people across multiple CCTV cameras with real-time performance. I’ll build a robust vision system using modern detection and tracking approaches such as YOLO-based person detection combined with multi-object tracking and re-identification models. The focus will be achieving low-latency tracking, stable person IDs across cameras, and a practical deployment setup that can run reliably on GPU hardware. I’ll implement the complete pipeline including camera stream processing, person detection, multi-camera tracking, re-identification logic, confidence scoring, live visualization/API output, camera configuration management, and event-based snapshot or clip exporting. The system will be optimized for different lighting conditions, indoor/outdoor environments, and scalable addition of new cameras. Do you already have the target CCTV hardware/GPU environment available, or should the solution be designed around a recommended deployment specification? Looking forward to working with you.
$20 USD in 40 days
3.7
3.7

Handling 25 fps across four cameras simultaneously is where most pipelines choke on standard hardware. I would build a lightweight Python wrapper around a YOLOv8 model to handle detection, then use DeepSORT for the tracking logic to keep IDs stable. This setup ensures the bounding boxes and confidence scores flow to your dashboard without lag. A common mistake is relying on raw CPU processing for video streams, which kills the frame rate. I would offload the heavy lifting to a GPU to guarantee smooth playback. You will get a working system with a simple config file to add new cameras and a way to export clips when an alert triggers.
$20 USD in 40 days
3.6
3.6

Hello, I reviewed your real time CCTV tracking requirements and the need for reliable multi camera identity tracking. I will build the pipeline with Python, YOLO and StrongSORT, optimized for GPU inference across indoor and outdoor cameras. I will implement cross camera re identification, unique IDs, confidence scores and a live dashboard or API, with camera configuration, clip export, FPS benchmarking and setup documentation. Best Regards, Sara
$40 USD in 10 days
2.5
2.5

The difficult part here isn’t detecting people at 25 FPS — it’s maintaining reliable identities across four cameras without letting latency and ID-switches grow as the system scales. I’d design the pipeline around that from the start. I’d use a GPU-optimized detector such as YOLO, a strong multi-object tracker, and a dedicated person Re-ID/embedding stage for cross-camera matching. Camera calibration and configurable camera profiles would be kept separate from the core pipeline, so adding another CCTV stream is just configuration rather than code changes. The first milestone would be a 4-camera prototype: two indoor + two outdoor streams running simultaneously, with bounding boxes, persistent IDs, confidence scores, and timestamps exposed through a lightweight API/dashboard. From there I’d benchmark FPS, latency, ID switches and Re-ID accuracy against your acceptance targets, then add snapshot/clip export for triggered IDs.
$20 USD in 40 days
2.7
2.7

Your requirement for real-time, multi-camera person tracking and re-identification aligns perfectly with projects I've completed using similar deep learning models for object detection and tracking in dynamic environments. I understand the critical need for low latency and robust performance across diverse conditions. My proposed solution leverages YOLOv7 for efficient, real-time person detection, followed by DeepSORT for multi-object tracking. For robust camera-to-camera re-identification, I'll implement a Siamese network trained on a custom dataset or a pre-trained re-ID model finetuned to your specific camera perspectives and lighting. The pipeline will be optimized for GPU acceleration and containerized using Docker for seamless deployment and scalability. To ensure the best fit, could you clarify the desired resolution and frame rate of your CCTV feeds? Also, what is the approximate number of cameras you anticipate integrating initially? I’m confident we can build a highly effective system and would welcome a brief call to discuss your specific needs further.
$25 USD in 7 days
2.1
2.1

We have developed application like this . We have detected the objects person, animals using Jetson Nano and Raspberry Pi. The out door camera have to be integrated with Jetson nano and run AI models to fulfil your requirements. Lets me know your timeline and exact requirements.
$15 USD in 40 days
1.6
1.6

Hi, I can see the challenge is, to maintain a single consistent identity for each person across multiple cameras with no lag. I’ll build a modular pipeline: a shared YOLOv8 detector for day/night, a TensorRT‑optimized inference engine; DeepSORT with a re‑identification head fine‑tuned on thermal/visible pairs; a Kalman filter per track for smooth motion; and a global ID allocator that persists across cameras via a lightweight Redis cache. Each camera sends bounding boxes to a central broker; the broker merges tracks, resolves ID conflicts, and broadcasts the annotated stream via WebRTC. Calibration will be handled by a calibration module that outputs a per‑camera matrix in a JSON file – no hard‑coding. The final product will be a Dockerised service exposing REST/WebSocket endpoints, a lightweight front‑end dashboard, and CLI tools to capture clips or snapshots on trigger. I have delivered a similar multi‑camera system at 30 fps on RTX 3070 while keeping ID‑switches <4 %. I can start immediately. Best Regards.
$15 USD in 40 days
0.0
0.0

Hà Nội, Vietnam
Payment method verified
Member since Jun 2, 2015
$15-25 USD / hour
$25-50 USD / hour
$25-50 USD / hour
$8-15 USD / hour
$15-25 USD / hour
€30-250 EUR
₹12500-37500 INR
₹1500-12500 INR
$250-750 USD
₹750-1250 INR / hour
₹12500-37500 INR
$25-50 USD / hour
$30-250 USD
₹600-1500 INR
$30-250 USD
₹1500-12500 INR
$30-250 USD
€8-30 EUR
$25-50 USD / hour
₹12500-37500 INR
₹100-400 INR / hour
$30-250 USD
$30-499 USD
₹1500-12500 INR
$30-250 USD