Számítógépes látás Jobs

I need a robust computer-vision pipeline that can automatically spot and label consumer-goods products in high-resolution images captured on our manufacturing line. The goal is to distinguish each finished item from background equipment, operators, and any other visual noise so that we can feed the detections into our downstream QA and inventory systems. You will start with raw JPEGs taken under factory lighting, then design, train, and validate an object-detection model—PyTorch or TensorFlow is fine—capable of achieving consistent, real-time performance on an NVIDIA GPU. If you prefer a different framework, let me know why; I’m flexible as long as the final solution is easy for my engineering team to maintain. Deliverables • Annotated sample dataset (in COCO or ...

We are looking to develop a solution capable of accurately detecting differences between two images. Project Description: We have: An original PNG file A scanned version of the same file after it has been printed The goal is to compare these two images and: Detect even the smallest differences Identify print-related defects (shifts, distortions, color variations, etc.) Perform detailed visual error analysis Requirements: High-precision or pixel-level comparison Robust handling of scan noise and imperfections Automatic image alignment (if needed) Output results as highlighted differences and/or a report Preferred Skills & Technologies: Python (OpenCV, Pillow, or similar libraries) Experience in Computer Vision and Image Processing Prior experience with similar projects is a plus...

I need a facial recognition system integrated into drones for secure delivery services. The system is essential for verifying identities and ensuring secure access to delivery points. Key Requirements: - Develop a robust facial recognition algorithm. - Integrate the system with drone technology. - Ensure real-time processing and accuracy. - Provide a user-friendly interface for monitoring. Ideal Skills and Experience: - Expertise in AI and machine learning. - Experience with drone technology. - Background in security systems. - Strong software development skills. Looking for candidates who can deliver ASAP.

I need a facial-recognition module that lets each of my delivery drones confirm a recipient’s identity before a package is released. The goal is simple: the aircraft hovers, points its camera at the person waiting below, and—within a fraction of a second—decides whether it’s safe to drop the parcel. Here’s what the system must do: • Detect and recognise faces in real time from a moving drone, outdoors, in daylight or at night. • Run on a light edge device carried by the drone (Jetson Xavier NX or similar) or stream efficiently to a ground server if you have a smarter approach. • Expose a clean API that returns the recognised ID, confidence score, and a yes/no flag my flight-control software can act on. • Allow me to enrol new authori...

I’m expanding an internal dataset used to train computer-vision models and need fresh, rights-cleared photos of urban landscapes that prominently feature buildings and architecture. Your own city—or any locale outside of our current archive—is perfect as long as the images are original and exclusively licensed to me. Focus of each shot • Clear view of architectural elements: façades, skylines, bridges, residential blocks, heritage sites and similar built structures. • Street vehicles or park spaces may appear incidentally, but the building itself must remain the dominant subject. What to deliver • High-resolution JPEG or PNG (minimum 2048 px on the shortest side; RAW files welcomed as an extra). • Simple filename noting the city or landmar...

I have a Unitree Go2 and need it to navigate outdoors without manual input. The focus is straight-forward: build and integrate a ROS 2 solution that lets the robot plan, localise, and execute safe paths across natural terrain. You are free to leverage the Nav2 stack, slam_toolbox, or any other ROS 2 packages that suit the task, as long as the end result runs reliably on the Go2’s onboard computer and works with its existing sensors. Deliverables • ROS 2 workspace with all custom nodes, launch files, and parameter sets • Clear build & deployment instructions (README or tutorial video) so I can reproduce the behaviour on a fresh system • Short live demo or rosbag proving the robot can start, plan, and reach a goal point over outdoor ground (grass/gravel) witho...

I have collected hundreds of standard photographs of hands in many poses and from many people. Nothing is uniform: skin tones, hand sizes, lighting, backgrounds, and even finger counts may vary. The goal is to turn this dataset into a practical analysis tool that can: • calculate the flexion/extension angle at every finger joint in each image • flag when a finger or phalanx is completely absent so that it can be recorded automatically I expect you will combine classical computer-vision preprocessing with a deep-learning model (PyTorch or TensorFlow are fine) and possibly leverage pose-estimation libraries such as MediaPipe or OpenCV for landmark detection. Accuracy matters more than perfection in lighting or background, so robust data-augmentation and domain-adaptation tech...

I have a growing archive of high-resolution intraoral photos and I need an AI mechanism that can automatically locate every tooth in each image, assign the correct FDI or Universal numbering, and clearly classify each tooth type (incisor, canine, premolar, molar, wisdom). You are free to decide whether a pre-trained or custom-built model is best; accuracy and speed matter more to me than the underlying brand of framework. Python with PyTorch or TensorFlow, OpenCV for preprocessing, and a clean REST or CLI inference interface would suit my workflow, but if you have a more elegant stack, I’m open. Key deliverables • A trained model capable of processing new intraoral 3d scans and returning JSON with tooth positions, numbers and classes. • Lightweight script or API e...

I need a complete, ready-to-run video analytics tool that flags car-collision accidents in city-surveillance footage. The workflow must stay simple for the end user: they upload a clip, hit “Analyze,” and, within seconds, the screen returns something like: Accident Detected: YES Severity: HIGH (low | medium | high) Confidence: 92 % Time-stamp: 00:12 Key points to build in • Scope of detection: only car collisions; no pedestrian or bicycle tracking at this stage. • Source footage: city CCTV style feeds (fixed street-level cameras). • No real-time push notifications are required—the result can appear once processing is finished. I will rely on you to select or curate a robust, publicly available dataset (or a combination of datasets) that...

# AI-Based Football Pitch Video Analytics System – Freelancer Project Brief ## Project Overview We are looking for a freelancer or team to develop an AI-powered computer vision and player performance analysis system for indoor football / mini football pitches. The system will be used in multiple football pitch locations across Türkiye. The software must: * Track every player during the match * Analyze player and team performance * Automatically generate statistics and ratings * Upload the match analysis to a website shortly after the match ends A complete hardware setup is already available for the first pitch: * 6 cameras installed * 1 high-performance PC * NVIDIA RTX 5080 GPU * Remote access via Tailscale is already configured The first implementation will be done on t...

I’m assembling a practical textbook that teaches drone technologies through hands-on programming with Python. The book should walk readers from fundamental concepts straight into real-time applications, letting them write, test, and fly code as they learn. Core coverage has to include basic Python syntax for newcomers, real-time data processing techniques, and solid drone control algorithms that work on popular flight controllers (PX4, ArduPilot, DJI SDKs, etc.). Example projects—think live telemetry dashboards, autonomous waypoint missions, and computer-vision-assisted flight—should be coded step-by-step so readers can replicate results on their own hardware. Because it’s a learning resource, clear explanations, commented code, troubleshooting tips, and small rev...

I need a rock-solid, real-time player tracking module for football matches that guarantees the ID assigned to each athlete at kick-off never changes until the final whistle. Right now, our OpenCV–TensorFlow–YOLO pipeline sometimes swaps or loses IDs when athletes overlap, leave the frame briefly, or the camera angle shifts, and that ruins every speed, distance, position, and heat-map metric we generate. Key requirements • Sport: football. • Camera setup: five or more synchronized feeds. • Existing stack: OpenCV, TensorFlow, YOLO – your solution must plug into this environment. What I expect 1. A multi-object tracker with integrated re-identification that preserves the same unique ID through occlusion, crossings, short disappearances, or camera changes...

Looking for an experienced Android Java developer to build a complete app for an AR glasses device. The device connects to an Android phone via USB-C. The app must: — Open a live USB UVC camera feed fullscreen on the glasses display — Apply real-time zoom from 1x to 10x with sharpness maintained — Apply real-time contrast enhancement — Apply real-time brightness adjustment — Apply edge detection outline mode — Apply colour filter modes — Read visible text aloud using OCR and text to speech — Simple on-screen menu to switch between features — Maintain minimum 30 FPS at all times SDKs provided: — — USB camera, zoom, image effects (Java API over C/C++ native libraries) — — glasses communication and display cont...

I have a set of GelSight images and need an end-to-end workflow that turns them into accurate, high-resolution 3D meshes of everyday objects. The goal is to capture subtle surface variations—scratches, fine textures, and edge curvature—so the final model can be inspected, measured, and exported for downstream use in CAD or visualization tools. You may choose whichever environment you are most comfortable with (Python + OpenCV/PyTorch, MATLAB, or C++). What matters is that the pipeline: • takes raw GelSight data as input (I will supply sample files), • performs all necessary photometric calibration and image alignment, • reconstructs a dense height map, then converts it to a watertight mesh, • outputs OBJ or STL alongside a brief report on reconstruct...

I need a clear, evidence-based report that compares the performance of today’s most widely cited object-detection algorithms. The focus is strictly on Computer Vision, zeroing in on Object Detection, and the core goal is to evaluate how the main approaches stack up against each other in terms of accuracy, speed, computational cost, and real-world suitability. Scope • Analyse at least three state-of-the-art methods—think Faster R-CNN, SSD, YOLO (v7/8), DETR or similar. • Draw all claims from peer-reviewed journals, top-tier conference papers, or authoritative benchmark leaderboards (e.g., COCO, PASCAL VOC). • Present metrics consistently (mAP, FPS, FLOPs, params, latency) so direct comparison is effortless. • Highlight strengths, weaknesses, and trad...

I need a clear, evidence-based report that compares the performance of today’s most widely cited object-detection algorithms. The focus is strictly on Computer Vision, zeroing in on Object Detection, and the core goal is to evaluate how the main approaches stack up against each other in terms of accuracy, speed, computational cost, and real-world suitability. Scope • Analyse at least three state-of-the-art methods—think Faster R-CNN, SSD, YOLO (v7/8), DETR or similar. • Draw all claims from peer-reviewed journals, top-tier conference papers, or authoritative benchmark leaderboards (e.g., COCO, PASCAL VOC). • Present metrics consistently (mAP, FPS, FLOPs, params, latency) so direct comparison is effortless. • Highlight strengths, weaknesses, and trad...

**Project Title:** Precision Extraction of Black Overprints from High-Resolution Stamp Sheets (Research Project) **Project Description:** I am conducting a specialized philatelic research study on 1931 Zeppelin overprints. I have two full stamp sheets (50 stamps each, high-resolution scans provided as PDF files), and I need the overprint isolated from each individual stamp with maximum fidelity. This is NOT a graphic design or basic background removal job. This is a precision image extraction task for research purposes requiring strict adherence to instructions and preservation of original detail. The source files are high-resolution PDF scans. You must extract images at native resolution (e.g., 1200 dpi). Screenshots, resampling, or any quality loss are not acceptable. If needed, conver...

I have just 700 indoor photos that must be labelled for an my project upcoming object-detection model. Every visible person and every seat that is currently occupied needs its own bounding box, and because the network will train on oriented-bounding-box (OBB) data, the rectangles have to follow the exact rotation of the body or seat—especially when someone is leaning. Data is of infrared camera of a cinema check samples , please note there can be many people in one image i want each to be marked properly You will work inside the simple browser-based annotation tool I built for this purpose or any other annotation tool you like. drag the box corners, spin the angle handle where needed, hit save, and move to the next frame; the app tracks progress automatically so nothing is missed....

AI Tool: Convert Architectural Casework Elevations (PDF) into Editable AutoCAD DWG Files with Self-Training Interface Project Description I need a custom AI tool that automatically reads architectural casework elevation drawings (PDFs) and generates complete, editable AutoCAD DWG files using my existing multiple block libraries. The tool must detect cabinets, sinks, fixtures, shelving, countertops, and miscellaneous equipment, then place the correct blocks with accurate dimensions, countertop outlines, and section views. It must also include a natural language self-training interface so a non-technical user can teach and refine the AI using plain English rules or by correcting the output drawing. This is Phase 1 (Proof of Concept) of a larger initiative. Future phases will expand to multip...

Hybrid Artificial Intelligence System for Autonomous Drone Detection Using Computer Vision and RF Signal Analysis The idea is to build AI detection system based on hybrid method (Computer vision and RF signal) and present results :Accuracy, Precision, Recall and False Alarm Rate Algorithms: Computer vision: YOLOv8 (Object Detection) RF analysis: CNN for Spectrogram analysis Merge in : Rule-Based Fusion • or Neural Network Fusion Dataset : Vision Datasets: • Anti-UAV Dataset • VisDrone RF Datasets: • DeepSig RadioML Dataset • Custom SDR Capture (HackRF / RTL-SDR) Budget : 150$

We are looking to hire experienced freelancers to help build a next-generation AI-powered camera management system for a camera business that is moving from traditional operations to a more advanced, centralized, and intelligent platform. The objective is to develop a scalable solution that can support multi-camera monitoring, AI-based body scanning, behavior tracking, vehicle number plate recognition, drone integrations, real-time alerts, and secure admin control. The platform should be designed for future growth and long-term operational use. This is not a basic prototype — we are looking for professionals who have worked on relevant, production-level systems. Project Scope The system will include, but is not limited to, the following capabilities: 1. Live camera streaming and re...

We are looking for a talented frontend developer to bring our Figma designs to life by building a visually stunning, high-performance website using HTML, CSS, and JavaScript. This project involves translating Japanese content, so proficiency in Japanese is required. The ideal candidate will have a strong eye for design and be able to accurately replicate layouts while enhancing them with smooth animations and professional styling. Responsibilities: Convert Figma designs into clean, responsive HTML/CSS/JavaScript code Ensure pixel-perfect implementation aligned with the original design Translate Japanese content accurately into the website Implement modern UI animations and interactions Optimize for performance and cross-browser compatibility Requirements: Proven experience in frontend deve...

I have just 700 indoor photos that must be labelled for an my project upcoming object-detection model. Every visible person and every seat that is currently occupied needs its own bounding box, and because the network will train on oriented-bounding-box (OBB) data, the rectangles have to follow the exact rotation of the body or seat—especially when someone is leaning. Data is of infrared camera of a cinema check samples , please note there can be many people in one image i want each to be marked properly You will work inside the simple browser-based annotation tool I built for this purpose or any other annotation tool you like. drag the box corners, spin the angle handle where needed, hit save, and move to the next frame; the app tracks progress automatically so nothing is missed....

We are hiring remote contributors to create photo-based language data using everyday materials found around you. This project focuses on collecting natural, real-life text captured through a phone camera. What You’ll Do - Photograph common objects that contain written text (printed or handwritten). - Provide three unique shots per item, changing position, distance, or lighting. - Ensure content is original and varied. - Most of the visible text (minimum 75%) must be in your local language. Eligibility - Fluent in the target language (native or near-native). - Physically located in a country where the language is used. - Own a smartphone capable of taking clear photos. How It Works - Upload images through a Google Form. - Submissions are reviewed individually. - Only valid, clear, ...