
Closed
Posted
Paid on delivery
I need a license-plate recognition engine that can reliably read Qatar, UAE, and Saudi Arabia plates when they appear in either still images or live video streams. High accuracy is non-negotiable—I am targeting production-level performance suitable for enforcement and analytics, so false positives must be minimal and confidence scores exposed. To succeed you will design, train, and package a full ANPR pipeline that: • detects the plate region in diverse lighting and angles, • recognises Arabic and Latin characters as well as the specific colour codes and icons used across the three countries, and • delivers results fast enough for real-time video (traffic-camera frame rates) while also accepting batch image inputs. Please outline your preferred tech stack (OpenCV, YOLOv8, TensorFlow, custom CNNs, etc.), any data you already hold for these plate formats, and what extra data or annotations you would need from me. Your deliverable should include the trained model, inference code (Python preferred), and a simple REST or gRPC interface so we can drop it straight into our existing pipeline. Acceptance will be based on: • ≥ 95 % character recognition accuracy on a blind test set I will provide, • ≤ 100 ms average inference per HD video frame on an NVIDIA T4 or equivalent, and • clear installation and usage documentation. If you have previously deployed ANPR systems for GCC plates—or can demonstrate equivalent multilingual OCR work—please share examples. I’m ready to move quickly once I see a solid approach.
Project ID: 40754721
58 proposals
Remote project
Active 5 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
58 freelancers are bidding on average $1,996 USD for this job

Hi There, Hope you are well. I specialize in developing high-accuracy license plate recognition engines for Qatar, UAE, and Saudi Arabia. My ANPR pipeline includes plate detection, character recognition (Arabic and Latin), and real-time processing. I will leverage OpenCV, TensorFlow, and custom CNNs for efficient results. I am prepared to meet your performance criteria and provide thorough documentation upon completion. Thanks in advance for the opportunity to working with you. Best Regards, Serazum
$2,500 USD in 1 day
7.6
7.6

Hello, I trust you're doing well. I am well experienced in machine learning algorithms, with nearly a decade of hands-on practice. My expertise lies in developing various artificial intelligence algorithms, including the one you require, using Python, and similar tools. I have worked with pytorch, and tensorflow to develop DL models, .I hold a doctorate from Tohoku University and have a number of publications in the same subject. My portfolio, which showcases my past work, is available for your review. Your project piqued my interest, and I would be delighted to be part of it. Let's connect to discuss in detail. Warm regards. please check my portfolio link: https://www.freelancer.com/u/sajjadtaghvaeifr
$1,850 USD in 7 days
7.3
7.3

Hi, I will deliver a Python ANPR engine that detects Qatar, UAE, Saudi plates in images and HD video, returns characters with confidence via REST API, meeting ≥95% accuracy and ≤100 ms inference on an NVIDIA T4. I can ship model and code in 4 weeks for $2,500. Ready to start; may I have a few annotated samples? Waiting for your response in chat! Best Regards.
$2,250 USD in 3 days
5.3
5.3

Your 95% accuracy target on GCC plates will fail if the model treats Arabic numerals (٠-٩) and Latin digits (0-9) as separate classes without contextual validation—most open-source ANPR pipelines collapse at the character-set boundary because they lack region-specific training data for Qatar's blue-stripe format, UAE's color-coded emirates, and Saudi's bilingual layouts. Quick questions - do you already have annotated training sets for all three plate formats, or will I need to source and label 5K+ images per country? And are you processing fixed-angle enforcement cameras or dashcam footage with variable pitch/yaw? Here is the architectural approach: - YOLOV8 + CUSTOM CNN: Two-stage pipeline—YOLOv8n for sub-20ms plate localization, then a character-segmentation CNN trained on synthetic GCC data augmented with real enforcement stills to handle glare and motion blur. - TENSORFLOW SERVING: Package the inference stack as a TensorFlow Serving container with gRPC endpoints, returning bounding boxes, character strings, confidence per glyph, and plate-type metadata (emirate color, Saudi region code). - BENCHMARK COMPLIANCE: I will deliver a Python test harness that logs per-frame latency on T4 hardware and computes precision/recall against your blind set, plus Dockerized deployment with CUDA 11.8 so your ops team can scale horizontally. I've built similar multilingual OCR for a Dubai tolling client that processes 12M plates monthly with 97.2% first-pass accuracy. Let's schedule a 20-minute technical call to review your camera specs and existing annotation workflow before I finalize the data-collection plan.
$2,030 USD in 30 days
5.5
5.5

I understand you require an ANPR engine with production-level accuracy for GCC plates, similar to the challenges in developing robust systems for various regional license plates where diverse character sets and environmental conditions are paramount. My experience includes building and optimizing object detection and character recognition models that have achieved high precision even with complex datasets. My approach will involve fine-tuning a state-of-the-art YOLOv8 or similar object detection model for plate localization, specifically trained on a curated dataset encompassing GCC plate variations, angles, and lighting. For character recognition, I'll leverage a transformer-based OCR model, adapted for both Arabic and Latin scripts, and incorporate custom layers to interpret color codes and icons. The pipeline will be containerized using Docker for seamless deployment and will expose confidence scores for each detection and recognition step. To ensure optimal performance, what specific confidence score threshold are you targeting for acceptable plate detections and character recognitions? Additionally, do you have any existing datasets or preferred data augmentation strategies that I should consider? I'm confident I can deliver a high-accuracy solution and would be happy to discuss your project in more detail.
$2,250 USD in 21 days
4.0
4.0

Hello, Deploying a high-accuracy Automatic Number Plate Recognition (ANPR) engine for GCC plates requires specialized deep learning pipelines to handle dual-script (Arabic/Latin) character sets, country-specific emblems, and varying aspect ratios under dynamic lighting. As a Computer Vision Engineer, I can build and package a real-time, production-grade ANPR system tailored to your requirements. Technical Architecture & Stack: Detection & Alignment: Deploy YOLOv8-pose fine-tuned on GCC plate geometries to perform 4-point perspective transformation, normalizing angled and skewed plates. Recognition & Multi-Head OCR: Train a custom CRNN (CNN + BiLSTM + CTC loss) or PaddleOCR-based model to read Arabic numerals, English text, and regional layout identifiers simultaneously. Optimization & Inference: Accelerate models using NVIDIA TensorRT / ONNX Runtime to achieve sub-50 ms inference per HD frame on an NVIDIA T4 GPU. Data Requirements & Deliverables: Data Strategy: Utilize existing GCC synthetic plate generator scripts and real-world datasets, requiring ~500–1,000 annotated regional sample images from you for final fine-tuning. Deliverables: Trained TensorRT/PyTorch model weights, modular Python inference script, low-latency FastAPI / gRPC microservice wrapper, Docker container, and complete deployment documentation. Ready to demonstrate past OCR pipelines and start immediately! Best regards, Collins
$2,250 USD in 7 days
5.1
5.1

Hi, I can develop a production-focused ANPR pipeline for Qatar, UAE, and Saudi plates, covering plate detection, Arabic/Latin OCR, country-specific formats, confidence scoring, still images, and real-time video. I’d recommend Python with OpenCV and a YOLO-based detector, combined with a specialized OCR pipeline that can be trained and evaluated specifically for the GCC plate formats and difficult lighting/angle conditions. I’ll optimize the inference pipeline for NVIDIA T4-class hardware and benchmark the complete detection-plus-recognition workflow against your 100 ms per HD-frame requirement. Your blind test set will be used for objective evaluation, with accuracy, false positives, confidence scores, and performance clearly reported before final delivery. I’ll provide the trained model, Python inference code, REST/gRPC interface, installation documentation, and a clear list of the additional plate images and annotations needed to reach the required production accuracy. Best regards, Paul Edward.
$2,000 USD in 7 days
4.1
4.1

Nice to meet you , My name is Anthony Muñoz, I express my interest in working on your project after carefully reading the requirements and concluding that they match my area of knowledge and skills. I am currently the lead engineer for the IT agency DSPro and I have more than 10 years of experience in the field. I have successfully completed a large number of similar jobs and I consider your project to be a challenge in which I would like to work and be able to make it a reality. Please feel free to contact me, it will be my pleasure to help you. I greatly appreciate the time provided and I remain attentive to any questions or concerns. Greetings
$1,974 USD in 7 days
3.8
3.8

Hello, Please provide the blind-test dataset, representative Qatar/UAE/Saudi plate samples, annotation format, and target video resolution/frame rate. I’ll use these to define the detection/OCR pipeline and benchmark requirements before training. I can build a production-focused ANPR pipeline using Python, OpenCV, deep-learning detection, and multilingual OCR, with plate localization, Arabic/Latin recognition, country-specific formatting, confidence scores, batch/video inference, and a REST API. I’ll benchmark accuracy and T4 latency, optimize inference, and provide the trained model, deployment code, and documentation. Best Regards, Sara..
$1,800 USD in 20 days
3.4
3.4

Building a GCC ANPR engine that hits 95% character accuracy on a blind test set is a data problem before it is a model problem, so I will be blunt about the plan. Detection: a YOLOv8-style detector trained for plates under diverse lighting and angles, since GCC footage varies widely in plate size, tilt and color schemes. Recognition: a multilingual OCR stack handling Arabic and Latin characters plus the country-specific color codes and icons. Both stages ship with confidence scores so your pipeline can route low-confidence reads to manual review instead of producing false positives. On honesty about data: I have built detection and recognition pipelines in PyTorch, but I do not hold a GCC plate dataset. The plan is to start from my detector and OCR baselines, fine-tune on annotated GCC plates you provide, and validate on your blind test set. With a few hundred annotated images, your target becomes achievable; without any GCC data, no model reaches it. Inference: TensorRT-exported model with a Python REST wrapper, benchmarked at 100ms or less per HD frame on a T4 or equivalent. Deliverables: trained weights, inference code, REST interface, installation and usage docs. https://www.freelancer.com/u/AlonzoLeeeooo
$2,025 USD in 21 days
3.6
3.6

For enforcement-grade accuracy on GCC plates, the biggest lever is something most ANPR engines ignore: Saudi and many UAE plates print the same information twice, in Arabic-Indic and in Latin characters (Saudi letters map one-to-one, e.g. أ = A). Reading both halves independently and only accepting a result when they agree turns most misreads into "low confidence" instead of a wrong plate, which is exactly how you keep false positives near zero. Qatar plates (digits only) and UAE emirate codes/colour bands get their own format rules, so a read that doesn't fit a valid pattern for its country is rejected. On video, results are tracked per vehicle across frames and voted, so one blurry frame can't produce a ticket. Pipeline: YOLO plate detector (robust to night/glare/angles), perspective correction, a character recogniser trained on real plus synthetic plates for rare formats, and a confidence score per character and per plate, exported to ONNX/TensorRT for real-time FPS, with a REST API for batch images. I've built YOLO-based traffic vision before (a closed-loop traffic-signal controller) and work in face/vision alignment. How many labelled plate images do you already have per country?
$1,500 USD in 12 days
2.3
2.3

I can develop a production-ready ANPR pipeline for Qatar, UAE, and Saudi plates using YOLO/OpenCV with custom Arabic + Latin OCR, plate-format/color/icon classification, confidence scoring, and GPU-optimized real-time inference. I’ll deliver the trained models, Python inference code, REST/gRPC API, batch and video processing, T4 performance optimization, blind-test validation, and complete installation/documentation.
$2,250 USD in 7 days
2.7
2.7

Hello, I understand you need a production-ready ANPR engine capable of recognizing Qatar, UAE, and Saudi Arabian license plates from images and live video. The goal is to deliver high-accuracy multilingual plate detection and OCR with confidence scores and real-time performance. Here’s what I can provide: Computer-vision pipeline using Python, OpenCV, YOLO, and deep-learning OCR for plate detection and Arabic/Latin character recognition. Support for country-specific plate layouts, colors, symbols, varied lighting, angles, and batch/live video inputs. Optimized inference with exposed confidence scores, trained model, Python inference code, REST/gRPC API, and complete documentation. I bring over 4+ years of experience in Python, computer vision, OpenCV, machine learning, and deep-learning applications, focusing on reliable and production-oriented solutions. Just to clarify a few things: Do you already have annotated Qatar, UAE, and Saudi plate datasets, or should data preparation/annotation be included? What camera resolution, FPS, and typical viewing distance/angle should the system be optimized for? Please come to the chat box to discuss more about your project. Best regards Indresh Kushwaha
$2,250 USD in 7 days
1.7
1.7

Hi, You need a high-throughput, dual-script (Arabic/Latin) ANPR pipeline optimized for GCC plate formats. Operationally, the system must localize plates in varying conditions from live streams, classify unique regional icons/color codes, perform high-confidence OCR, and expose parsed results via a REST/gRPC endpoint without exceeding the 100ms inference budget on an NVIDIA T4. Technical approach: - Plate Localization: YOLOv8 optimized for diverse lighting and camera angles. - OCR & Feature Classification: A customized CRNN for handling right-to-left Arabic and Latin scripts concurrently, plus a lightweight CNN for color/icon recognition. - Optimization: Exporting models to TensorRT/ONNX to guarantee sub-100ms execution on the T4 GPU. - Delivery: Containerized Python application (Docker) exposing a gRPC/REST API for streaming and batch ingestion. Core modules: - Stream ingestion and batch processing handler. - Two-stage inference engine (YOLOv8 localization -> Cropping -> Multi-script OCR). - Confidence scoring and JSON payload generator. Relevant systems: - Alpha Robot (Edge CV and AI hardware integration) - Bison360 ERP (Real-time scanning and inventory automation) Implementation strategy: We will start by aggregating an initial GCC plate dataset, followed by training the YOLOv8 detector. After validating localization, we will train the OCR models, compile to TensorRT, benchmark on a T4, and deliver the API for your blind testing. Regards, Rohit
$1,500 USD in 21 days
0.8
0.8

Hi, I build computer-vision pipelines in Python and deploy them as fast APIs on GPU. One point on acceptance: • 95% on your blind set depends on training data that matches it. Public GCC plate data is thin, so I'll need 3,000-5,000 labelled plates per country, or we label your footage together. I'll report accuracy per country so problems are visible early. Plan: 1. Detection: YOLOv8/YOLO11 for plates, trained on day/night, angles, motion blur. 2. Recognition: CRNN/PARSeq OCR for Arabic + Latin characters, plus a classifier for colour codes, categories and emirate/country. 3. Rules: per-country format checks to cut false reads; confidence score per character and plate. 4. Speed: export to TensorRT FP16, target under 100 ms per HD frame on a T4; tracking (ByteTrack) so one car = one read. 5. FastAPI REST (gRPC optional), Docker image, test report on your blind set. Questions: 1. How many images/hours of video do you have per country, and are they labelled? 2. Camera type: fixed traffic camera or mobile? Resolution and FPS? 3. Edge device or server GPU in production? 4. Do you need motorcycle and temporary plates? Timeline: 40-45 days including data prep, training and testing. Price: $2,800, a production model with measured accuracy, not a demo on clean images. Tanvir
$2,800 USD in 45 days
0.6
0.6

This project demands a robust ANPR engine capable of high accuracy in recognizing license plates from Qatar, UAE, and Saudi Arabia. My approach would begin with designing a custom CNN using TensorFlow, supplemented by OpenCV for image processing to handle various lighting and angles. Once the model is trained and optimized to meet the target accuracy, I will package it with a REST API for seamless integration. I possess hands-on experience in similar domains, including AI integration, and can ensure performance within your specified time frame. I would need clarification on the data you currently have for training, especially regarding annotations and examples, to tailor the solution effectively. For reference, I developed an AI-based platform that showcases my capability in real-time processing and machine learning functions. What existing data do you have for training the ANPR model?
$1,500 USD in 14 days
0.0
0.0

Hello, I checked your project "High-Accuracy GCC ANPR Engine" and already have a clear understanding of your requirements and how to deliver them efficiently. I have solid experience in Python, and I've successfully completed similar projects by delivering high-quality, scalable, and reliable solutions. Why choose me? * Strong expertise in Python * Clean, optimized, and scalable implementation * Fast communication and regular progress updates * Focused on delivering results, not just completing the task I'm ready to start immediately and would be happy to discuss the details. Best regards, Umer
$1,500 USD in 3 days
0.0
0.0

First paragraph in Arabic recognition is not where this project starts: it starts with the dataset gap, since you have not yet confirmed what annotated plate imagery exists for Qatar, UAE, and Saudi formats, and that gap drives both the 95 percent target and the 100ms budget. Delivery: a YOLOv8-based detector for plate region plus a CRNN/CTC recognition stage trained per-country to separate Arabic script, Latin transliteration, and the colour/icon codes, exposed via a FastAPI REST service returning confidence scores, packaged with Docker and a README. v1, a working detector plus draft recognition on whatever sample images you provide, within 7 working days. Here is the REST response contract I would implement first: {"plate": "DXB-A-12345", "country": "UAE", "confidence": 0.97, "bbox": [112,340,220,60], "frame_ms": 42} First two decisions: use YOLOv8 for detection since it handles angle/lighting variance well on small objects, and train three separate recognition heads rather than one shared model, since character sets and plate layouts differ enough across the three countries to hurt accuracy if merged. I do not promise 95 percent accuracy before seeing real annotated samples from your blind test distribution. Can you share a sample batch of labelled plates per country now. Price: 1500 USD fixed · Timeline: 12 working days from access/brief Konrad
$1,500 USD in 12 days
0.0
0.0

A production-ready enforcement system that eliminates manual verification and delivers high-confidence GCC plate data in real-time. To achieve ≥95% accuracy across Qatar, UAE, and Saudi Arabia, I will implement a pipeline using YOLOv8 for high-speed plate localization and a specialized CRNN (Convolutional Recurrent Neural Network) with an attention mechanism to handle the dual Arabic-Latin character sets and regional icons. To meet the <100ms latency requirement on an NVIDIA T4, I will optimize the final model using TensorRT for maximum inference throughput. I will handle the entire process: - Design and training of the regional detection and OCR models. - Implementation of a high-performance Python inference engine. - Packaging as a gRPC/REST interface for seamless integration into your pipeline. Within the first 5 days, I will deliver a baseline prototype demonstrating detection and recognition on a sample of your GCC plate data so you can verify the trajectory. I've successfully deployed similar computer vision pipelines requiring strict latency and accuracy constraints. I'll ensure the confidence scores are exposed to allow your system to flag low-certainty reads automatic
$3,000 USD in 21 days
5.1
5.1

Hey, I can build an ANPR engine for Qatar, UAE, and Saudi plates, accounting for differences in plate layouts, Arabic and Latin text, and image conditions. Do you already have labeled plate images from all three countries? The available data and its variety will determine whether the best route is fine-tuning a model or building a pipeline that combines detection and OCR. I’ll deliver a tested engine that detects plates and returns recognized text with confidence scores. How I would run it: 1. Review the data and define evaluation metrics for each country. 2. Build and test the detection and recognition pipeline, including image preprocessing and text normalization. 3. Handover: source code, setup instructions, and country-level test results. I’ll track accuracy separately by country and check common failure cases such as glare, blur, and low light. Plate-layout variation and OCR edge cases are central to the approach, not afterthoughts. What format should the engine accept and return, and where will it run? Aakash P.S. I’ll include a sample-based validation plan so accuracy can be measured clearly. Just send a "Message" so we can talk :)
$1,500 USD in 14 days
0.0
0.0

Guntur, India
Member since Mar 17, 2021
$750-1500 USD
$250-750 USD
$250-750 USD
₹37500-75000 INR
₹100-400 INR / hour
₹1500-12500 INR
$15-25 USD / hour
₹750-1250 INR / hour
₹750-1250 INR / hour
₹12500-37500 INR
₹1000-3000 INR
€30-250 EUR
$30-250 USD
$15-25 USD / hour
$30-250 AUD
₹600-1500 INR
$2-8 USD / hour
₹750-1250 INR / hour
$250-750 USD
$2-8 USD / hour
₹37500-75000 INR
$99-100 USD
₹12500-37500 INR