You will get GYM Trainer - Squat Analysis, Pull-Up Analysis
Project details
Turn any phone camera into an AI personal trainer. I build computer vision fitness apps that count reps, detect exercise form errors, and give real-time coaching feedback — no wearables, no depth camera, no gym equipment sensors.
Using pose estimation and custom-trained models, the system tracks joint angles frame by frame to judge whether a squat hit depth, whether a pull-up reached full extension, and whether a deadlift is putting the user's back at risk.
What you get:
✅ Rep counting with high accuracy on standard phone video
✅ Form scoring tuned to your coaching standard, not a generic default
✅ Real-time on-device feedback — no upload lag
✅ Workout history, rep logs, and progress tracking
✅ Up to 10 exercises: squats, push-ups, pull-ups, deadlifts, lunges, curls, planks and more
✅ Full source code, model weights, and documentation — you own it outright
Built for: fitness app startups, online coaching platforms, gym chains, physical therapy and rehab tracking, sports training programs.
Computer Vision Engineer at Ultralytics (creators of YOLO), 30+ CV projects delivered. Send me your exercise list and a sample clip — I'll tell you honestly what's achievable before you buy.
Using pose estimation and custom-trained models, the system tracks joint angles frame by frame to judge whether a squat hit depth, whether a pull-up reached full extension, and whether a deadlift is putting the user's back at risk.
What you get:
✅ Rep counting with high accuracy on standard phone video
✅ Form scoring tuned to your coaching standard, not a generic default
✅ Real-time on-device feedback — no upload lag
✅ Workout history, rep logs, and progress tracking
✅ Up to 10 exercises: squats, push-ups, pull-ups, deadlifts, lunges, curls, planks and more
✅ Full source code, model weights, and documentation — you own it outright
Built for: fitness app startups, online coaching platforms, gym chains, physical therapy and rehab tracking, sports training programs.
Computer Vision Engineer at Ultralytics (creators of YOLO), 30+ CV projects delivered. Send me your exercise list and a sample clip — I'll tell you honestly what's achievable before you buy.
Machine Learning Tools
Deeplearning4j, Keras, NumPy, NVIDIA AI Platform, OpenCV, Python, PyTorch, scikit-learn, TensorFlowWhat's included
| Service Tiers |
Starter
$300
|
Standard
$350
|
Advanced
$1,500
|
|---|---|---|---|
| Delivery Time | 3 days | 3 days | 7 days |
Number of Revisions | 1 | 1 | Unlimited |
Number of Model Variations | 1 | 1 | 5 |
Number of Graphs/Charts | 4 | 4 | 20 |
Model Validation/Testing | |||
Model Documentation | |||
Data Source Connectivity | - | - | |
Source Code | - | - |
Optional add-ons
You can add these on the next page.
Fast Delivery
+$100 - $500Frequently asked questions
About Ammar
Computer vision engineer and deep learning expert and AI specialist
Islamabad, Pakistan - 12:11 am local time
Hi! I’m Ammar Ali, a Computer Vision Engineer with 2.5+ years of experience building production-grade AI and deep learning solutions for real-world applications.
I specialize in Computer Vision, Object Detection, Multi-Object Tracking, Image Segmentation, Pose Estimation, Edge AI, and AI model optimization.
🚀 What I Can Build for You
✅ YOLOv8 / YOLOv10 / YOLOv11 Object Detection
✅ Object Tracking — DeepSORT, ByteTrack, SORT, MOT
✅ Image & Video Segmentation — SAM-2, Mask R-CNN
✅ Pose Estimation & Keypoint Detection
✅ OCR & Computer Vision Automation
✅ Real-Time Video Analytics & Surveillance
✅ Vehicle, Person & Traffic Detection
✅ Custom AI Model Training & Fine-Tuning
✅ Model Optimization — Quantization, Pruning, Distillation
✅ ONNX & TensorRT Optimization
✅ NVIDIA Jetson Edge AI Deployment
✅ FastAPI / Flask AI APIs
✅ Docker, Kubernetes & MLOps
✅ AWS, Azure & GCP Deployment
🛠️ Tech Stack
Python | PyTorch | TensorFlow | Keras | OpenCV | YOLO | SAM-2 | ONNX | TensorRT | CUDA | FastAPI | Flask | Docker | Kubernetes | MLflow | AWS | Azure | GCP
📈 Experience Highlights
- Built real-time smart-city and surveillance vision systems.
- Optimized AI inference on NVIDIA Jetson devices using TensorRT and INT8 quantization.
- Reduced inference latency by 67% in an edge deployment project.
- Developed multi-object tracking and real-time detection pipelines.
- Built scalable video inference APIs using FastAPI and distributed GPU workers.
- Developed vehicle detection and parking-management solutions for 500+ parking spaces.
- Worked on production AI systems from model training and testing to deployment and monitoring.
💡 Why Work With Me?
I don't just train models — I focus on building complete, production-ready AI solutions that are accurate, fast, scalable, and deployable.
If you have a Computer Vision or AI project, send me your requirements and let's build it together.
Steps for completing your project
After purchasing the project, send requirements so Ammar can start the project.
Delivery time starts when Ammar receives requirements from you.
Ammar works on your project following the steps below.
Revisions may occur after the delivery date.
Requirements & Exercise List
We lock the exercise list, target platform (iOS, Android, web, or edge device), and what "correct form" means for each movement. You send sample footage of the conditions the app will actually run in. I confirm feasibility before any code is written.
Dataset Collection & Annotation
I gather and label video data for each of the 10 exercises — rep boundaries, joint keypoints, form-error cases. If you have proprietary footage or a specific coaching standard, this is where it gets encoded into the model.