shahin.fit estimates body composition from a single full-body photo plus your height and weight, then turns the result into a clear, personal health plan. No calipers, no gym scanner, no appointment.
Two people with the same weight and BMI can have very different body fat and very different health risks. Accurate measurement today means DEXA scans, bioimpedance devices or a trained person with calipers: expensive, inconvenient, and rarely repeated often enough to track progress.
It can't tell muscle from fat, so it misclassifies athletes and misses people with normal weight but high body fat.
Clinical body composition scans cost money and time, so most people measure once, if ever.
Even when people get a number, they rarely get a clear explanation of what it means or what to do next.
Stand straight in front of a plain wall. Take one front-facing, full-body photo and enter your height and weight.
Computer vision finds 33 body landmarks and measures shoulder, waist, hip, torso and leg proportions, scaled to your real height.
Built and working todayThe measurements are sent to our analysis service, which uses the Claude API to estimate body composition and reason about the result in context.
In developmentYou get a plain-language report, a realistic target range, and a personal nutrition and training plan. Re-scan every few weeks to track progress.
The first prototype ran entirely on a laptop: MediaPipe pose detection fed hand-crafted body measurements into a small neural network (ANN) trained on a public 252-person body-fat reference dataset. It proved the pipeline works. It also showed the limits of a small, self-trained model, so the analysis layer is moving to the Claude API.
# analysis service (simplified) import json from anthropic import Anthropic client = Anthropic() report = client.messages.create( model="claude-sonnet-5-5", max_tokens=1500, system="You are a body composition analyst. Return JSON: body_fat_range, category, explanation, plan.", messages=[{"role": "user", "content": json.dumps({ "height_cm": 178, "weight_kg": 98, "age": 27, "landmark_ratios": {"shoulder_hip": 1.77, "waist_hip": 1.38, "leg_torso": 1.36}, "goal": "lose fat, keep muscle", })}], )
Each image below was processed by our pose pipeline. Green lines mark the shoulder, waist and hip widths used for measurement.
Labels show shoulder/hip and leg/torso ratios measured by the prototype. Reference images are synthetic test images, not real users.
Estimated body-fat range, category and what it means for you, explained in plain language.
Calorie and protein targets, weekly training structure and realistic milestones based on your goal.
Re-scan every few weeks. See how your proportions and estimates change, and ask the AI coach why.
shahin.fit is at the very beginning. The core vision pipeline works; the next step is replacing the local model with an API-based analysis engine and putting it in front of real users.
Pose detection, measurement extraction and a local neural network for body-fat estimation.
Backend service, structured analysis and report generation through the Claude API; evaluation against reference measurements.
Upload, analyze and receive a report in the browser. First 100 testers.
On-device capture with pose guidance, progress history and the AI coach.
Accuracy study against bioimpedance/DEXA measurements; pilots with gyms and dietitians.
Trust is the product. These principles are part of the design from day one.
We're looking for beta testers, fitness professionals and partners. Write to us and we'll get back to you.
shn@shahin.fitFounded by Shahin Bayramov.