컴퓨터 비전 Jobs

I’m building an AI-driven solution that keeps single-family homes healthy by spotting structural issues, plumbing problems, and electrical faults while the house is in everyday use. The goal is a system that can take inputs such as periodic smartphone photos, CCTV streams, or smart-sensor readings and flag likely defects in real time so owners receive clear, actionable alerts before small flaws turn into costly repairs. Here is what I need from you: • A trained model (or ensemble) able to classify and localise those three defect categories with high precision on typical residential imagery or sensor data. • A lightweight inference pipeline that can run on consumer-grade hardware or be exposed through a cloud API. • A simple web or mobile interface where a homeow...

The goal is to deliver a stand-alone Python desktop application that records classroom or workplace attendance through real-time face recognition. The stack is fixed: Python 3, OpenCV, and the standard Tkinter GUI toolkit. Core flow 1. A compact Tkinter window opens with clearly labelled buttons. The interface must, at a minimum, let a supervisor register new users. 2. During registration the system captures multiple webcam frames per person, automatically saving at least 20 crisp face crops for stronger training data. 3. An LBPH model is trained on those images and stored locally; subsequent launches should reload the latest model without retraining. 4. When recognition mode is active the app checks each detected face against the model and writes a time-stamped entry to a CSV fil...

I run a production Python pipeline that extracts structured data from complex technical PDF documents — detecting objects, tracing linear features, and measuring areas and regions — using a mix of vision LLMs and computer vision. It works and is deployed, but I need someone who can take real ownership and push accuracy, robustness, and cost in directions I can't get to fast enough on my own. The pipeline is staged: render the PDF, read the document's reference key to learn what to look for, detect every instance on each page with a spatial vision model, measure each one (OpenCV contour/skeleton tracing, with an LLM agent fallback), and hand off structured results. What I need help with: - Improving detection accuracy and recall on dense, messy,...

I want to deploy an end-to-end AI agent that oversees my entire plastics line—starting at the raw-material warehouse and finishing on the pallet truck. Scope The system must handle three core domains: 1. Quality control • Real-time defect detection on every part coming off the machines • Continuous consistency checks against dimensional tolerances and colour specs • Material-compliance verification that flags any recipe or lot that drifts outside specification 2. Production optimisation • Extrusion, injection-moulding and thermoforming lines each need parameter-tuning models that minimise cycle time and scrap while maintaining quality • Predictive maintenance recommendations based on machine signals and historical downtime ...

I need a reliable, camera-based solution that can automatically detect and count people captured by our existing CCTV network, regardless of whether the footage comes from indoor halls or outdoor entrances. The end goal is clear: turn those raw counts into actionable customer-analytics data I can slice by time, location, and traffic trend. Here’s what I’m expecting: • Accurate real-time or near-real-time counts from standard CCTV streams (no extra infrared hardware). • Robust performance under varying light and weather because some of our cameras sit outdoors. • An easy way for me to view and export daily, weekly, and monthly traffic reports—CSV download and a lightweight dashboard are perfect. • A straightforward deployment path: containerised code (...

I need a reliable pipeline that takes any pre-recorded video, swaps the on-screen face with a chosen target face, and keeps the lips perfectly synced to the original soundtrack. High accuracy is non-negotiable—phoneme-level alignment, natural mouth shapes, and seamless facial blending should hold up under close inspection and slow-motion playback. Preferred stack You’re free to combine proven solutions such as Wav2Lip, DeepFaceLab, FaceSwap, or custom GAN models, as long as the end result meets the visual standard. Python with PyTorch/TensorFlow is ideal because I want the option to retrain or fine-tune later, but an all-in-one executable is acceptable if thoroughly documented. Required deliverables • Use e.g. Kling AI, Wan 2.3 Runway ML Acceptance criteria •...

I’m looking for an experienced researcher–author to co-develop, write, and successfully submit a skin-disease–detection paper to a genuine, APC-free SCI-indexed journal and takes acceptance guarantee. Scope • Craft a manuscript that introduces and validates a brand-new detection algorithm. • Build experiments around whichever deep-learning model is currently most promising—CNN, transformer, hybrid, or any other trending architecture—so long as the results outperform or clearly benchmark against existing methods. • Handle all stages from dataset curation, model training, result analysis, and figure/table preparation through to journal formatting, cover letter, and submission. • Guide the paper through peer review until acceptance. ...

Project Overview: We are building a data analysis platform and require a Computer Vision / Digital Signal Processing expert to implement a server-side automated motion analysis engine. The frontend application layer is already complete and successfully uploads a cropped "eye-strip" video (focusing strictly on the eye region) directly to our server backend. Your Task: you will leverage existing open-source frameworks (such as OpenCV contour tracking, MediaPipe Iris, or similar established eye-tracking libraries) to build a highly reliable server-side script (Python preferred). This script must trigger automatically upon video upload to process the cropped video frame-by-frame. Core Technical Deliverables: 1. Object Tracking & Noise Filtering: Utilize MediaPipe Iris or OpenCV...

A new cross-disciplinary build is about to start, combining solid software engineering with practical artificial-intelligence features. The first milestone centres on scoping: together we will decide whether the core product emerges as a web, mobile or desktop application and lock down the primary language stack—Python, JavaScript, Java or a hybrid approach. From there, the work shifts to architecture design, clean code implementation, and the integration of an AI component that demonstrably adds value (machine-learning, NLP, or computer-vision models will be chosen once use-cases are finalised). Version control through Git, test coverage, and concise technical documentation are expected throughout the sprint. The immediate deliverable is a working proof of concept that showcases a...

I have a collection of images featuring plates that contain both numbers and alphabet characters. The plates are already labeled, but I need each annotation carefully validated and, whenever necessary, corrected. Your task is to open every image, check that the plate region truly matches the ground-truth label, and fix any inaccuracies you see so the dataset is production-ready. Accuracy is absolutely critical here—if a bounding box is off by even a few pixels or if a character has been mis-read, the file must be updated. I would like the final result to be a clean set of corrected annotation files (JSON or XML, whichever you prefer) plus the revised images, zipped and ready for model training. A brief change log noting the number of fixes you made will also help me keep track of ov...

I have a complete OCT (optical coherence tomography) image set and now need to turn it into a full, publication-ready study on age-related macular degeneration (AMD). The work has to run entirely in Google Colab and revolve around a hybrid deep-learning architecture of your choice—CNN + transformer, ensemble CNNs, or any comparable combination—as long as it meets strong SCI journal standards. Pre-processing Both FFT and Wavelet Transform must be applied. Please document each step in the notebook so the signal-processing pipeline is clear and reproducible. Core modelling • Train, validate and test the model on the OCT data. • Track and store all metrics so they can be plotted later. • Incorporate Explainable AI focused on feature-importance visualisation...

I have a single 40-second MP4 clip that shows two motorcycles circulating the same track. Each bike can be separated at a glance because they are painted different colours. What I need is a reliable, frame-accurate measurement of the time interval between the first and the second motorcycle as they pass a chosen reference line on the circuit. Please use YOLO (or an equivalent real-time object detector) to: • detect both bikes throughout the whole sequence, • define a consistent reference line or region on the track, • timestamp the exact moment each bike crosses that reference, and • burn a clear visual overlay onto the video that displays the calculated gap in seconds. The finished deliverable is the processed video with the overlay already embedded; no separa...

I am in the middle of my master’s thesis and need a capable partner to handle the coding side of a deep-learning project on mammogram images. The core task is to build a robust model for breast-cancer diagnosis; once the network architecture is in place I can take care of statistical write-ups and broader discussion, but I need you to get the codebase production-ready and well-documented. Here’s what the collaboration will look like: • Data: I will supply the curated mammography dataset and any preprocessing scripts I have so far. • Model building: Using Python with either TensorFlow or PyTorch (whichever you work fastest with), you will design, train and fine-tune a CNN or transformer-based pipeline that targets breast-cancer detection specifically. • Resu...

I need to turn a Large Language Model into a practical, real-time assistant that runs directly on an NVIDIA Jetson board. The model will reason over image data at the edge, so every millisecond saved and every megabyte spared matters. Memory-optimised execution is therefore the single most important constraint, though I still expect you to keep latency low and power draw sensible. Here is what I want to achieve. The LLM must accept a pre-processed visual input, apply prompt-engineering tricks that preserve context, and reply fast enough to be useful on-device—no cloud fallback. Smart caching, selective quantisation, and, when it genuinely pays off, lightweight fine-tuning are all on the table. I will look to you to suggest the right mix of techniques and to implement them. You shou...

I need a lean, working proof-of-concept that automatically counts foot traffic using a single 360-degree camera. The goal is to drop the unit into busy conference halls, festival entrances, or outdoor promotional zones and have it return reliable head-counts without manual intervention. Here is what matters to me: • Vision logic: Please build or integrate computer-vision models (OpenCV, YOLO, TensorFlow Lite or similar) that detect and track people moving through the camera’s full 360° field of view. The algorithm must distinguish unique passes so that every person is counted once. • Edge or cloud flexibility: I am fine with the model running on a Raspberry Pi 4, Jetson Nano, or a small cloud instance—as long as latency is low and setup remains simple. • ...

We are Sprectex AI, a growing artificial intelligence company specializing in Computer Vision, AI automation, and enterprise-level AI solutions. We have a strong team of 50+ AI engineers and specialists across different regions, actively working on real-world deployments and client projects. Our focus is on building scalable, production-ready AI systems for global clients. We already have multiple working AI systems deployed with real clients, and we are now expanding our operations into the US market. We are currently looking for a serious and experienced US-based Business Development Partner / Sales Representative who can help us generate leads, connect with enterprise clients, and close high-value deals. The ideal candidate will be responsible for identifying potential clients, buil...

I have a complete OCT (optical coherence tomography) image set and now need to turn it into a full, publication-ready study on age-related macular degeneration (AMD). The work has to run entirely in Google Colab and revolve around a hybrid deep-learning architecture of your choice—CNN + transformer, ensemble CNNs, or any comparable combination—as long as it meets strong SCI journal standards. Pre-processing Both FFT and Wavelet Transform must be applied. Please document each step in the notebook so the signal-processing pipeline is clear and reproducible. Core modelling • Train, validate and test the model on the OCT data. • Track and store all metrics so they can be plotted later. • Incorporate Explainable AI focused on feature-importance visualisation...