File size: 2,429 Bytes
e7be713 b178e9b f3d0069 30e9b4f f926004 30e9b4f f3d0069 30e9b4f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 | import streamlit as st
import cv2
import numpy as np
import mediapipe as mp
from PIL import Image
st.title("👕 Virtual Try-On with Pose Estimation")
person_file = st.file_uploader("Upload Person Image", type=["jpg", "jpeg", "png"])
cloth_file = st.file_uploader("Upload Clothing Image (PNG)", type=["png"])
mp_pose = mp.solutions.pose
pose = mp_pose.Pose(static_image_mode=True)
def overlay_cloth(person_img, cloth_img):
img_rgb = cv2.cvtColor(person_img, cv2.COLOR_BGR2RGB)
result = pose.process(img_rgb)
if not result.pose_landmarks:
st.warning("Pose not detected.")
return person_img
h, w, _ = person_img.shape
landmarks = result.pose_landmarks.landmark
left_shoulder = landmarks[mp_pose.PoseLandmark.LEFT_SHOULDER]
right_shoulder = landmarks[mp_pose.PoseLandmark.RIGHT_SHOULDER]
left_hip = landmarks[mp_pose.PoseLandmark.LEFT_HIP]
right_hip = landmarks[mp_pose.PoseLandmark.RIGHT_HIP]
# Shoulder and hip coordinates
x1, y1 = int(left_shoulder.x * w), int(left_shoulder.y * h)
x2, y2 = int(right_shoulder.x * w), int(right_shoulder.y * h)
x3, y3 = int(left_hip.x * w), int(left_hip.y * h)
x4, y4 = int(right_hip.x * w), int(right_hip.y * h)
# Width and height based on shoulder width and upper body height
cloth_width = int(np.linalg.norm([x2 - x1, y2 - y1]) * 1.2)
cloth_height = int(np.linalg.norm([((x3 + x4) // 2) - ((x1 + x2) // 2), ((y3 + y4) // 2) - ((y1 + y2) // 2)]) * 1.3)
# Center the clothing at chest
center_x = (x1 + x2) // 2
center_y = (y1 + y2) // 2
# Resize cloth
cloth_img_resized = cloth_img.resize((cloth_width, cloth_height), Image.Resampling.LANCZOS)
# Compute top-left corner for overlay
paste_x = int(center_x - cloth_width // 2)
paste_y = int(center_y - cloth_height // 3)
# Convert person_img to PIL for blending
person_pil = Image.fromarray(cv2.cvtColor(person_img, cv2.COLOR_BGR2RGB)).convert("RGBA")
cloth_img_resized = cloth_img_resized.convert("RGBA")
# Paste with transparency
person_pil.paste(cloth_img_resized, (paste_x, paste_y), cloth_img_resized)
return np.array(person_pil)
if person_file and cloth_file:
person_img = np.array(Image.open(person_file).convert("RGB"))
cloth_img = Image.open(cloth_file)
result_img = overlay_cloth(person_img, cloth_img)
st.image(result_img, caption="Result", use_column_width=True)
|