The problem Harmony AI solves

Our project provides an AI-powered solution for overall wellness, featuring real-time SOS detection through discrete hand gestures for personal safety, instant skin allergy diagnosis from uploaded images for accessible health insights, and a supportive chatbot for mental health assistance, offering a holistic approach to enhance safety, health, and emotional well-being

Challenges we ran into

Challenges We Ran Into

SOS Gesture Detection Sensitivity

Challenge: Detecting the specific SOS hand gesture sequence (open hand, tuck thumb, trap thumb, and release) was challenging. Variations in hand position, incomplete gestures, or minor movements led to false positives and unreliable detection.
Solution: We used MediaPipe Hands to detect and track hand landmarks and implemented a sequential tracking system. This approach required each step of the gesture to be completed in the correct order within a specific time frame, significantly reducing errors.
Handling Variability in Skin Allergy Images

Challenge: The AI model for skin allergy detection struggled with inconsistent image quality due to differences in lighting, skin tones, and image resolution. These factors affected the model’s ability to make accurate predictions.
Solution: We applied image normalization techniques to adjust for lighting conditions and augmented our dataset with diverse images to improve robustness. By standardizing preprocessing, we ensured the model was less impacted by variations in image quality.
Creating an Empathetic Chatbot for Mental Health

Challenge: Designing a chatbot that responds empathetically and provides relevant support for mental health was difficult. Generic responses and insensitivity could potentially make the bot ineffective or even counterproductive for users in distress.
Solution: We fine-tuned our chatbot model on mental health support datasets, focusing on empathetic dialogue. Additionally, we implemented response filtering to reduce generic answers, improving the quality and sensitivity of interactions on topics like loneliness and depression.
Maintaining Real-Time Performance for SOS Detection

Challenge: Real-time gesture detection requires rapid processing to be effective in emergencies. The system initially struggled with processing speed, causing delays in recognizing the SOS gesture..