FitVice
Fitvice is an AI-powered fitness app offering personalized workouts, smart nutrition, live coaching, posture tracking, and mental wellness tools—all in one seamless, intelligent platform.
Created on 20th April 2025
•
FitVice
Fitvice is an AI-powered fitness app offering personalized workouts, smart nutrition, live coaching, posture tracking, and mental wellness tools—all in one seamless, intelligent platform.
The problem FitVice solves
🔹 Personalized Workouts Made Easy
Fitvice generates AI-tailored workout plans based on your fitness level, goals, and progress—no need to scroll through generic YouTube routines.
🔹 Real-Time Posture Correction
Using computer vision, Posture Sense™ tracks your movements, counts reps, and corrects your form instantly—reducing injury risk and improving performance.
🔹 Smart Nutrition, Simplified
Skip the calorie math. Fitvice’s AI nutrition guide and recipe generator suggest balanced meals aligned with your fitness goals, dietary needs, and schedule.
🔹 Holistic Health in One App
From BMI/weight tracking to guided meditation, yoga sessions, and expert consultations—Fitvice eliminates the clutter of juggling multiple wellness apps.
🔹 Muscle-pedia: Train Smarter
Access detailed exercise guides with video demos to ensure you’re training the right muscle, the right way, every time.
🔹 Stay Organized, Stay Motivated
A built-in fitness to-do list helps you track daily goals, build habits, and stay consistent—keeping your health journey on track.
Challenges we ran into
Hurdle: Real-Time Posture Detection Accuracy
One of the biggest challenges we faced was implementing real-time posture analysis using computer vision. The issue was that movement recognition was inconsistent across different lighting conditions and body types, leading to inaccurate rep counting and poor form correction feedback.
How We Solved It:
Optimized the model: We fine-tuned a lightweight pose estimation model (MoveNet) to balance speed and accuracy for mobile responsiveness.
Data augmentation: Added diverse body types and lighting scenarios to the training dataset to improve generalization.
Confidence thresholding: Implemented a confidence threshold to filter out low-quality pose detections and reduce false positives.
Fallback mechanisms: For edge cases, we added a manual rep-counting toggle as a backup to ensure user progress wasn't lost
Tracks Applied (1)
Groq track
Groq