|
|
|
|
|
import gradio as gr |
|
|
import cv2 |
|
|
import numpy as np |
|
|
from PIL import Image |
|
|
import mediapipe as mp |
|
|
import torch |
|
|
from torchvision import transforms |
|
|
from torchvision.models import resnet18 |
|
|
import torch.nn.functional as F |
|
|
import openai |
|
|
import os |
|
|
|
|
|
|
|
|
openai.api_key = os.getenv("OPENAI_API_KEY") |
|
|
|
|
|
|
|
|
def generate_smart_plan(wake_time, swim_time, swim_duration, gym_today, sleep_hours, goal): |
|
|
plan = [] |
|
|
if sleep_hours < 7: |
|
|
plan.append("⚠️ Less than 7 hours of sleep — prioritize recovery or a nap.") |
|
|
plan.append(f"🌅 Wake-up time: {wake_time}") |
|
|
plan.append(f"🏊 Swim: {swim_time} for {swim_duration} minutes — Focus: {goal.capitalize()}") |
|
|
if gym_today: |
|
|
if goal == "speed": |
|
|
plan.append("🏋️ Gym Plan: Explosive lifts — med ball slams, box jumps.") |
|
|
elif goal == "endurance": |
|
|
plan.append("🏋️ Gym Plan: Circuits with light weights and high reps.") |
|
|
else: |
|
|
plan.append("🏋️ Gym Plan: Core, mobility, and injury prevention.") |
|
|
else: |
|
|
plan.append("💪 No gym — light stretching and mobility recommended.") |
|
|
plan.append("🍽️ Eat protein + carbs within 30 minutes post-swim.") |
|
|
plan.append("🌙 Target 8–9 hours of sleep tonight.") |
|
|
\ |
|
|
timeline = [] |
|
|
timeline.append(f"⏰ {wake_time} - Wake Up + Hydrate") |
|
|
try: |
|
|
swim_hour, swim_min = map(int, swim_time.split(":")) |
|
|
swim_end_hour = swim_hour + swim_duration // 60 |
|
|
swim_end_min = (swim_min + swim_duration % 60) % 60 |
|
|
timeline.append(f"🏊 {swim_time} - Swim Practice ({swim_duration} min)") |
|
|
timeline.append(f"🥤 {swim_end_hour:02}:{swim_end_min:02} - Post-swim Nutrition") |
|
|
except: |
|
|
timeline.append("❓ Unable to parse swim time — add manually to your calendar.") |
|
|
|
|
|
timeline.append("💪 Gym: Today" if gym_today else "🧘 Active Recovery: Mobility/Stretching") |
|
|
timeline.append("🛌 Bedtime: Aim for 8–9 hours of sleep") |
|
|
|
|
|
return "\n".join(plan) + "\n\n🕒 Sample Timeline:\n" + "\n".join(timeline) |
|
|
|
|
|
|
|
|
food_classes = ["pizza", "salad", "burger", "sushi", "spaghetti", "steak", "pancakes"] |
|
|
food_macros = { |
|
|
"pizza": {"cal": 285, "protein": 12, "carbs": 36, "fat": 10}, |
|
|
"salad": {"cal": 150, "protein": 5, "carbs": 10, "fat": 7}, |
|
|
"burger": {"cal": 500, "protein": 25, "carbs": 40, "fat": 30}, |
|
|
"sushi": {"cal": 200, "protein": 10, "carbs": 30, "fat": 5}, |
|
|
"spaghetti": {"cal": 350, "protein": 12, "carbs": 60, "fat": 8}, |
|
|
"steak": {"cal": 400, "protein": 40, "carbs": 0, "fat": 25}, |
|
|
"pancakes": {"cal": 350, "protein": 6, "carbs": 45, "fat": 12} |
|
|
} |
|
|
|
|
|
model = resnet18(pretrained=True) |
|
|
model.eval() |
|
|
transform = transforms.Compose([ |
|
|
transforms.Resize((224, 224)), |
|
|
transforms.ToTensor() |
|
|
]) |
|
|
|
|
|
def predict_macros_with_explanation(img): |
|
|
image = Image.fromarray(img) |
|
|
x = transform(image).unsqueeze(0) |
|
|
with torch.no_grad(): |
|
|
logits = model(x) |
|
|
probs = torch.nn.functional.softmax(logits, dim=1).squeeze().numpy() |
|
|
pred_idx = np.argmax(probs) |
|
|
food = food_classes[pred_idx % len(food_classes)] |
|
|
macros = food_macros.get(food, {"cal": 0, "protein": 0, "carbs": 0, "fat": 0}) |
|
|
|
|
|
base = f"🍽️ Predicted: {food}\nCalories: {macros['cal']} kcal\nProtein: {macros['protein']}g\nCarbs: {macros['carbs']}g\nFat: {macros['fat']}g" |
|
|
|
|
|
try: |
|
|
response = openai.ChatCompletion.create( |
|
|
model="gpt-3.5-turbo", |
|
|
messages=[ |
|
|
{"role": "system", "content": "You are a helpful sports nutritionist."}, |
|
|
{"role": "user", "content": f"This food is {food} with macros: {macros}. Give a short explanation on how this helps a swimmer recover after training."} |
|
|
] |
|
|
) |
|
|
gpt_reply = response['choices'][0]['message']['content'] |
|
|
except Exception as e: |
|
|
gpt_reply = f"⚠️ GPT error: {str(e)}" |
|
|
|
|
|
return base + "\n\n💡 GPT Insight:\n" + gpt_reply |
|
|
|
|
|
|
|
|
mp_pose = mp.solutions.pose |
|
|
|
|
|
def analyze_swim_video_with_gpt(video_path): |
|
|
cap = cv2.VideoCapture(video_path) |
|
|
pose = mp_pose.Pose() |
|
|
issues_detected = set() |
|
|
frame_count = 0 |
|
|
|
|
|
while cap.isOpened(): |
|
|
ret, frame = cap.read() |
|
|
if not ret or frame_count > 60: |
|
|
break |
|
|
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) |
|
|
results = pose.process(frame_rgb) |
|
|
if results.pose_landmarks: |
|
|
l_shoulder = results.pose_landmarks.landmark[mp_pose.PoseLandmark.LEFT_SHOULDER] |
|
|
r_shoulder = results.pose_landmarks.landmark[mp_pose.PoseLandmark.RIGHT_SHOULDER] |
|
|
l_hand = results.pose_landmarks.landmark[mp_pose.PoseLandmark.LEFT_WRIST] |
|
|
r_hand = results.pose_landmarks.landmark[mp_pose.PoseLandmark.RIGHT_WRIST] |
|
|
if abs(l_hand.y - l_shoulder.y) < 0.1: |
|
|
issues_detected.add("Left hand enters too flat") |
|
|
if abs(r_hand.y - r_shoulder.y) < 0.1: |
|
|
issues_detected.add("Right hand enters too flat") |
|
|
frame_count += 1 |
|
|
|
|
|
cap.release() |
|
|
|
|
|
if not issues_detected: |
|
|
return "✅ Stroke technique looks great in the first 60 frames!" |
|
|
|
|
|
tips = "\n".join(f"- {issue}" for issue in issues_detected) |
|
|
|
|
|
try: |
|
|
gpt_response = openai.ChatCompletion.create( |
|
|
model="gpt-3.5-turbo", |
|
|
messages=[ |
|
|
{"role": "system", "content": "You are a professional swim coach."}, |
|
|
{"role": "user", "content": f"The following stroke issues were detected: {tips}. Explain their impact on performance and how to correct them."} |
|
|
] |
|
|
) |
|
|
gpt_advice = gpt_response['choices'][0]['message']['content'] |
|
|
except Exception as e: |
|
|
gpt_advice = f"⚠️ GPT error: {str(e)}" |
|
|
|
|
|
return "⚠️ Technique Tips:\n" + tips + "\n\n💡 GPT Advice:\n" + gpt_advice |
|
|
|
|
|
|
|
|
profile_data = {} |
|
|
|
|
|
def save_profile(name, age, gender, goal, diet_type, pb100free): |
|
|
profile_data["name"] = name |
|
|
profile_data["age"] = age |
|
|
profile_data["gender"] = gender |
|
|
profile_data["goal"] = goal |
|
|
profile_data["diet"] = diet_type |
|
|
profile_data["pb100free"] = pb100free |
|
|
return f"✅ Profile saved for {name} (100 Free PB: {pb100free} sec)" |
|
|
|
|
|
|
|
|
daily_plan = gr.Interface( |
|
|
fn=generate_smart_plan, |
|
|
inputs=[ |
|
|
gr.Textbox(label="Wake-up Time (e.g. 5:30 AM)"), |
|
|
gr.Textbox(label="Swim Time (e.g. 6:00 AM)"), |
|
|
gr.Slider(label="Swim Duration (min)", minimum=0, maximum=180), |
|
|
gr.Checkbox(label="Did you go to the gym today?"), |
|
|
gr.Slider(label="Sleep Last Night (hrs)", minimum=4, maximum=10, step=0.5), |
|
|
gr.Dropdown(["speed", "endurance", "technique", "recovery"], label="Today's Focus") |
|
|
], |
|
|
outputs=gr.Textbox(label="AI Daily Plan"), |
|
|
title="📅 Daily Plan Generator" |
|
|
) |
|
|
|
|
|
macro_analyzer = gr.Interface( |
|
|
fn=predict_macros_with_explanation, |
|
|
inputs=gr.Image(type="numpy", label="Upload Food Image"), |
|
|
outputs="text", |
|
|
title="🥗 Macro Estimator + GPT Insight" |
|
|
) |
|
|
|
|
|
swim_analyzer = gr.Interface( |
|
|
fn=analyze_swim_video_with_gpt, |
|
|
inputs=gr.Video(label="Upload Swim Video"), |
|
|
outputs="text", |
|
|
title="🏊 Swim Analyzer + GPT Advice" |
|
|
) |
|
|
|
|
|
profile = gr.Interface( |
|
|
fn=save_profile, |
|
|
inputs=[ |
|
|
gr.Textbox(label="Name"), |
|
|
gr.Number(label="Age"), |
|
|
gr.Dropdown(["Male", "Female", "Other"], label="Gender"), |
|
|
gr.Dropdown(["speed", "endurance", "technique", "recovery"], label="Main Goal"), |
|
|
gr.Dropdown(["None", "Vegetarian", "Vegan", "High Protein"], label="Diet Type"), |
|
|
gr.Number(label="100 Free Personal Best (sec)") |
|
|
], |
|
|
outputs="text", |
|
|
title="👤 Swimmer Profile" |
|
|
) |
|
|
|
|
|
|
|
|
gr.TabbedInterface( |
|
|
[daily_plan, macro_analyzer, swim_analyzer, profile], |
|
|
tab_names=["📅 Daily Plan", "🥗 Macros + GPT", "🏊 Technique + GPT", "👤 Profile"] |
|
|
).launch() |
|
|
|
