AI Computer Vision Analysis MVP Development

@MashoodurRehman1

I am a seasoned ML/AI engineer with proficiency in Python and extensive experience in computer vision. Over the years, I have developed a deep understanding of video processing, OpenCV, and pose estimation algorithms, which aligns with the core requirements of your project. I am well-versed in utilizing deep learning frameworks like PyTorch and TensorFlow for designing efficient, modular analysis systems. Previously, I have built similar applications using facial recognition and activity classification methodologies. My experience with GPU optimization ensures efficient running of models on cloud or local platforms. I understand the nuances of identifying individuals using facial embeddings, classifying a limited set of exercises, and leveraging pose estimation for accurate rep counting, providing structured outputs as detailed in your requirements. I am interested in discussing how I can leverage my expertise in computer vision for your MVP. Could you share which exercise poses are prioritized for counting? I look forward to collaborating on this innovative project.

@AwaisChaudhry

Hello, I will build a compact Python pipeline to analyze pre-recorded workout videos and output who is in the clip, the exercise label, rep counts, and per-rep timestamps. The plan is to keep face id separate from exercise and rep logic for a clean, modular flow: (1) face identification using provided enrollment images, (2) pose-based exercise classification among a small set, and (3) rep counting via robust pose heuristics. The MVP will run locally or on a GPU-enabled environment and output structured results plus a simple README with run instructions. I will keep the scope tight for demonstration and board review, with a clear separation between components and sample outputs on your videos. What is the exact set of 5–8 exercises you want the MVP to recognize and classify? Questions I would ask you to nail the scope: What is the exact small set of exercises to recognize? How many enrollment images per known person will you provide? Do you want confidence scores included in the output? If yes, format preferred. What frame rate and resolution are typical in your videos? Any lighting or camera movement challenges we should assume? Which facial embedding model or library would you like to use (if any) for identity? Should the system handle partial occlusion or missing frames in a graceful way? Do you require any privacy or data handling notes for enrollment data? What is the preferred output format (JSON, CSV, or both) and sample schema?

@neerajdhameliya

Hello, Building a reliable fitness analytics MVP requires moving from "academic" models to robust, heuristic-driven logic that handles human movement variability. I can deliver a modular Python pipeline that turns raw workout footage into structured, board-ready data within your 4–6 week window. Technical Approach: For this internal demo, I prioritize stability and clarity over complex, black-box architectures: Face Identification: Lightweight pipeline using InsightFace or FaceNet embeddings, with cosine similarity for high-confidence identification in a closed set. Action Classification: Temporal segment networks or sliding-window approaches on MediaPipe/RTMPose coordinates. This keeps the system performant and easy to debug. Modular Architecture: Separate processors for Face, Pose, and Logic allow your team to swap models or update exercises without a full rewrite. Repetition Counting – Squats: Keypoint Normalization: Scale coordinates by user height and hip-width to handle camera distance. Angular Heuristics: Track hip, knee, and ankle angles; a rep starts when knee angle drops below 100° and counts when it returns above 160°. State Machine: Neutral, Descending, Peak, and Ascending states filter out partial reps and capture timestamps at peak flexion. To optimize the pose backbone, will videos be primarily profile (side) or frontal view? Best regards, Niral

@sajjadtaghvaeifr

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 Matlab, Python, and similar tools. 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

@divumanocha

As a seasoned professional who's been providing technical expertise for over a decade, I'm confident that I possess the ideal skill set to develop an exceptional AI computer vision analysis system for your workout video project. My proficiency in Python and deep understanding of computer vision fundamentals will allow me to flawlessly interpret and implement your requirements. Moreover, my wealth of experience with video processing, OpenCV, pose estimation as well as deep learning frameworks like PyTorch and TensorFlow, perfectly aligns with your needs. Not only am I capable of building a Python-based pipeline that processes pre-recorded workout videos to yield structured analysis results based on identified user, exercise label, rep count, per-rep timestamps, and confidence scores, but I can also design clean, modular systems which ensure optimum performance. Additionally, my knowledge of facial recognition or embedding-based identity systems could give a valuable edge to the project. My approach to the pose-based repetition counting technique is primarily twofold: identifying the specific joint movements indicative of a squat's starting and ending position and then quantifying the number of repetitions by monitoring these movements throughout the video Thanks....

@leadconcept

Greetings, I understand you are building a tightly scoped computer vision MVP to analyze pre recorded workout videos, identify known individuals, classify exercises, and count repetitions using pose estimation, with clean structured output for internal demonstration. Before we proceed, I would like to clarify a few things: 1, Do you have a preferred pose backbone, for example MediaPipe or a PyTorch based model? 2, Should exercise classification rely purely on pose sequences or combine visual features as well? 3, What GPU environment will be used for development and testing? Our team includes Python engineers experienced in OpenCV, pose estimation, facial embeddings, and modular deep learning pipelines using PyTorch. We design clean separations between identity, activity classification, and rep logic, with reproducible scripts and structured JSON outputs. Let us connect to review sample videos and environment details. The current bid amount is a placeholder to start the conversation. Regards Yasir LEADconcept PS: Let me know, if you want to see our team past work to determine our skills/expertise or past customer's references.

@devendrathakur12

Hello, I specialize in computer vision systems and built & customized large scale video analysis pipelines in Python. The main challenge here is keeping identity, exercise label, and rep counting clean and modular without overcomplicating the MVP. I am certified in Python and PyTorch development, and I will solve this by using OpenCV for video handling, face embeddings for closed-set ID, a small action classifier, and pose estimation (MediaPipe or a PyTorch model) for rep counting. For squats, I track hip and knee angles over time, detect full down-up cycles, and count clean motion peaks. A few questions: Are videos fixed camera angle? What frame rate and resolution? How strict should confidence filtering be? Will lighting vary a lot? Happy to share similar CV work. Best regards, Dev S.

@corpmember29

Hello, {{{ I HAVE CREATED SIMILAR APPS BEFORE AND I CAN SHOW YOU }}} I can deliver a clean, Python-based computer vision MVP that analyzes pre-recorded workout videos exactly within your stated scope. The pipeline will be modular, separating face identification (closed-set embeddings), exercise classification, and pose-based repetition counting. For squat rep counting, I will use pose estimation to track key joints (hip, knee, ankle), compute joint angles over time, apply temporal smoothing, and detect full down–up cycles with confidence scoring and per-rep timestamps. Outputs will be structured (JSON/CSV) with identified user, exercise label, rep count, timestamps, and confidence values. The solution will be GPU-friendly, reproducible, and focused on demonstrable accuracy rather than production hardening. I HAVE 10+ YEARS OF EXPERIENCE IN PYTHON, COMPUTER VISION, OPENCV, POSE ESTIMATION, AND DEEP LEARNING FRAMEWORKS (PYTORCH / TENSORFLOW), AND THIS PROJECT ALIGNS DIRECTLY WITH MY PRIOR WORK IN VIDEO ANALYSIS AND APPLIED ML SYSTEMS. I WILL PROVIDE 2 YEAR FREE ONGING SUPPORT AND COMPLETE SOURC CODE, WE WILL WORK WITH AGILE METHODOLOGY AND WILL GIVE YOU ASSISTANCE FROM ZERO TO PUBLISHING ON STOIRES. I eagerly await your positive response. Thanks.

@denisSolution

Hi client, I'm Denis Redzepovic, an experienced developer with expertise in OpenCV, Data Analysis, Deep Learning, Facial Recognition, Python, Pytorch, Video Processing, Computer Vision and Machine Learning (ML). I have worked extensively on diverse Python projects, ranging from backend development and automation to data processing and API integrations. My deep understanding of Python’s libraries and frameworks allows me to build efficient, scalable, and maintainable solutions. I pay close attention to code quality and performance to ensure your project runs flawlessly. With my solid experience, I’m confident I can deliver results that exceed your expectations. I focus on writing clean, maintainable, and scalable code because I know the difference between 99% and 100%. If you hire me, I’ll do my best until you’re completely satisfied with the result. Let’s discuss your project details so I can tailor the perfect Python solution for you. Thanks, Denis

@quan9824

https://www.freelancer.com/projects/raspberry-pi/Powered-Monitoring-Prototype-Development/reviews https://www.freelancer.com/projects/computer-vision/Real-Time-Windows-Face-Recognition/reviews && Computer Vision, OCR, OpenCV, Tensorflow, PyTorch, Keras, ML/DL model && Hi, How are you?. I have full skills and full experiences of this field. I have developed many Compuiter Vision systems and I am expert in these fields I can finish your project with high quality and on time. Please send me your message to discuss more about your project. I am waiting your reply now. Thanks.

@OutsourceMan

⭐⭐⭐⭐⭐ We at CnELIndia, led by Raman Ladhani, can deliver this MVP efficiently by leveraging our deep expertise in computer vision and Python-based video analytics. Our approach will involve building a modular pipeline with clear separation of face identification, exercise classification, and rep counting. For pose-based squat repetition counting, we will extract key joint coordinates (hips, knees, ankles) using a pre-trained pose estimation model, track their vertical displacement across frames, and detect full movement cycles with threshold-based heuristics. Facial embeddings will enable closed-set recognition of provided users, while a lightweight classifier will label exercises. Outputs will include user ID, exercise, rep count, per-rep timestamps, and confidence scores in a structured format. Raman’s prior experience in human activity recognition and GPU-friendly deployment ensures rapid prototyping and reproducible results. Sample videos and a concise README will accompany the deliverables.

@Rekhathakur

With a decade of experience under our belt, we at WellSpring Infotech are confident in our ability to tackle your applied engineering project in computer vision for your workout analysis MVP. Our python programming skills are robust and we have an impressive history of implementing intricate solutions such as the face identification you are looking for. Our familiarity with deep learning frameworks including PyTorch and TensorFlow will surely come in handy for this project, as well as our previous work in activity recognition. Regarding pose-based repetition counting, our team will employ our profound understanding of computer vision combined with state-of-the-art pose estimation techniques to evaluate and enumerate the squatting repetitions accurately. We recognize the value of modular system designs and will ensure clear separation between face identification, exercise classification, and rep counting to deliver structured results as required by you. Not only are we equipped with the technical skills crucial for your project, but we also understand the importance of maintaining an uncompromised level of quality throughout. Thanks.....

@sasaprogramer21

I can deliver this MVP as a clean Python pipeline that runs on prerecorded workout clips and outputs structured results for identity, exercise label, and repetition counting with timestamps and confidence. I have built similar internal demos that combine face embeddings, pose estimation, and heuristic rep counting with modular components and reproducible outputs. Architecture would be three separable stages. Face identification using closed set enrollment images to compute embeddings and match per clip with a confidence threshold. Exercise classification using pose driven features from a fixed exercise set, starting with lightweight temporal features and a simple classifier to stay reliable for an MVP. Rep counting as a dedicated module for at least squats with per rep timestamps and quality checks. Squat rep counting approach. Track hip knee ankle keypoints per frame and compute a smoothed knee angle and hip vertical displacement normalized by body scale. Define a state machine with top and bottom phases using thresholds and hysteresis. A rep starts when the subject transitions from top to descent and completes when returning to top after reaching a valid bottom depth. I can share relevant CV examples as code snippets and demo clips upon request, and I am ready to start February 23 with 20 to 30 hours per week. Estimated time: 4 to 6 weeks

@efanntyo

Hi there, I’m an experienced Python developer and computer vision engineer ready to deliver a compact, scalable MVP for your AI Computer Vision Analysis pipeline. My background spans OpenCV-driven video processing, pose estimation, facial embeddings, and deep learning in PyTorch , aligned with your requirements for a clean, modular Python solution that’s GPU-friendly and suitable for internal demonstrations. What I will deliver is a tightly scoped pipeline that processes pre-recorded workout videos to produce structured results including: identified user (via facial embeddings against your enrollment set), exercise label from a small fixed taxonomy, rep counts (for at least one exercise like squats) using robust pose-based heuristics, and per-rep timestamps with confidence scores. The codebase will clearly separate face identification, exercise classification, and rep counting logic, with a lightweight runner and an easily extensible configuration to add new exercises or users. How I’d approach the project: - Data & setup: accept enrollment images and a curated video set, define a small exercise taxonomy, and establish a common frame rate handling strategy. - Face identification: implement embedding-based recognition (e.g., using a lightweight face-embedding model) with a simple closed-set matcher against enrolled identities; return user labels or unknown when not matched. - Pose estimation: integrate a robust, real-time-friendly pose estimator (e.g., MediaPipe or a PyTor

@xuezhe315

Hello, As an experienced software engineer with a strong background in computer vision and machine learning, I am confident that I have the right blend of skills to deliver this AI Computer Vision Analysis MVP for you. Throughout my career, I have consistently shown my ability to design clean and modular systems which are essential for this project's requirements. My profound proficiency with Python and deep learning frameworks such as PyTorch combined with my expertise in OpenCV align seamlessly with the technical stack you're looking for. When approaching pose-based repetition counting for a squat using video, one aspect that's critical is accurate pose estimation. Over the years, I've worked extensively on various projects involving pose estimation and have developed effective heuristics that can substantially enhance rep-counting accuracy. Understanding of how minor variations on pose can be detected and counted as repetitions will be key. In addition to the proficiency you need, I also bring to the table my business-driven mindset, scalable solutions, and transparent communication which have been praised by my previous clients. This will ensure that we remain aligned on objectives from concept to production and deliver not only a functioning system but one that aligns closely with your internal goals is well within scope and is also flexible enough for future iterations. Let me turn your complex vision into a flawless reality! Best Regards.

@vinod00k

Leveraging my extensive experience in not only the realms of computer vision and Python but also deep learning frameworks like PyTorch and TensorFlow, I am more than capable for delivering on your AI Computer Vision Analysis MVP. I have successfully executed various projects using these skills, including creating apps over diverse industries that cater to real-world needs - which exactly aligns with your intents. My proficiency in video processing and OpenCV makes me suitable to tackle the complex aspects of identifying individuals in video streams as well as classifying exercises from a broad range of available ones. My understanding of pose estimation wouldn't just be operational but would also lend itself to implementing the required rep counting process for squats with an appreciable level of accuracy. Moreover, I pride myself on my ability to design clean, modular systems and build powerful yet easy-to-maintain pipelines. For this project, I would ensure high-quality code separated into distinct modules (face identification, exercise classification, and rep counting). Completing tasks within an optimal timeline without sacrificing quality is one of my major strengths which could bring immeasurable value to your tight timeframe. Let's create something extraordinary together!

@toriquldev123

Hello Dear! I write to introduce myself. I'm Engineer Toriqul Islam. I was born and grew up in Bangladesh. I speak and write in English like native people. I am a B.S.C. Engineer of Computer Science & Engineering. I completed my graduation from Rajshahi University of Engineering & Technology ( RUET). I love to work on Web Design & Development project. Web Design & development: I am a full-stack web developer with more than 10 years of experience. My design Approach is Always Modern and simple, which attracts people towards it. I have built websites for a wide variety of industries. I have worked with a lot of companies and built astonishing websites. All Clients have good reviews about me. Client Satisfaction is my first Priority. Technologies We Use: Custom Websites Development Using ======>Full Stack Development. 1. HTML5 2. CSS3 3. Bootstrap4 4. jQuery 5. JavaScript 6. Angular JS 7. React JS 8. Node JS 9. WordPress 10. PHP 11. Ruby on Rails 12. MYSQL 13. Laravel 14. .Net 15. CodeIgniter 16. React Native 17. SQL / MySQL 18. Mobile app development 19. Python 20. MongoDB What you'll get? • Fully Responsive Website on All Devices • Reusable Components • Quick response • Clean, tested and documented code • Completely met deadlines and requirements • Clear communication You are cordially welcome to discuss your project. Thank You! Best Regards, Toriqul Islam

@vishal1145

Your rep counting logic will fail if you rely on simple angle thresholds - squats have depth variations, camera angles shift, and occlusion happens when users rotate. I've debugged this exact failure mode in 3 fitness apps where "rep 1 = 10 reps" because the heuristic couldn't distinguish partial reps from full range of motion. Quick question - are your enrollment videos shot in the same lighting and camera setup as the workout clips? Face embeddings degrade fast when training data doesn't match inference conditions. Also, what's your tolerance for false negatives on exercise classification - would you rather the system say "unknown" or risk mislabeling a lunge as a squat? Here's the architectural approach: - PYTORCH + MEDIAPIPE: Use MediaPipe Pose for real-time keypoint extraction (33 landmarks at 30fps), then build a temporal smoothing layer using Kalman filters to handle jitter. This prevents false triggers when hip/knee angles fluctuate during transitions. - FACIAL RECOGNITION: Implement ArcFace embeddings with cosine similarity matching. I'll set up a two-stage pipeline - MTCNN for face detection, then embedding comparison with a 0.6 threshold to balance precision/recall. If your enrollment set is under 20 people, this runs in <50ms per frame on CPU. - EXERCISE CLASSIFICATION: Train a lightweight LSTM on pose sequences (not single frames) to capture movement patterns. I'll use a 2-second sliding window with 80/20 train/val split on your labeled data. This handles the transition frames between exercises that trip up image classifiers. - REP COUNTING STATE MACHINE: For squats, I track hip-to-knee angle and vertical displacement of the hip keypoint. A rep registers when: (1) hip drops below knee threshold, (2) hip rises back above starting position, (3) dwell time at bottom exceeds 200ms to filter bouncing. I'll log per-rep depth and tempo so you can flag partial reps in the output JSON. - VIDEO PROCESSING: Use OpenCV with frame skipping (process every 3rd frame) and batch inference to hit 15fps on a T4 GPU. Output includes frame-level annotations and a summary CSV with user_id, exercise_label, rep_count, timestamps, and confidence scores. I've built pose estimation pipelines for 2 sports analytics platforms that processed 50K+ videos. The failure modes are always the same - occlusion, lighting shifts, and edge cases like users stepping out of frame mid-rep. Let's schedule a 15-minute call to walk through your sample videos and discuss how to handle these scenarios before I start coding.

@mkh559a5a1fd66e9

With a notable focus on the Python programming language for over five years, I am no stranger to the challenges that machine learning and computer vision pose - in fact, I thrive on them. My exposure to OpenCV, deep learning frameworks like PyTorch and TensorFlow, and practical experiences in object detection makes me an ideal candidate for your AI Computer Vision Analysis MVP Development project. My focus on building clean, modular systems aligns well with your project requirements. For pose-based repetition counting of a squat using video, I would leverage my knowledge of pose estimation to create heuristics within the pipeline that recognize key posture changes characterizing a completed squat repetition. Through this solution, I will ensure that identified users' information, exercise label, rep count, per-rep timestamps, and confidence scores are all clearly separated- just as you've specified. Though I lack experience in fitness or biomechanics specifically (which would certainly be a great value-add), my understanding of mathematics and sharp problem-solving abilities equip me with the essential foundation to comprehend kinematic principles involved in human movements. Moreover, my familiarity with GPU inference and model optimization could further enhance the overall performance of the system. Trust me to build you a powerful yet sober product that will fulfill your unique needs.