File size: 3,189 Bytes
26fd668
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
efebb94
 
 
 
 
 
 
 
26fd668
efebb94
 
 
26fd668
 
efebb94
 
 
26fd668
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
import streamlit as st
import torch
import cv2
import mediapipe as mp
from transformers import SwinForImageClassification, AutoFeatureExtractor
from PIL import Image
import numpy as np

# Initialize face detection
mp_face_detection = mp.solutions.face_detection.FaceDetection(
    model_selection=1, min_detection_confidence=0.5)

# Initialize model and labels
@st.cache_resource
def load_model():
    id2label = {0: 'Heart', 1: 'Oblong', 2: 'Oval', 3: 'Round', 4: 'Square'}
    label2id = {v: k for k, v in id2label.items()}
    
    model = SwinForImageClassification.from_pretrained(
        "microsoft/swin-tiny-patch4-window7-224",
        label2id=label2id,
        id2label=id2label,
        ignore_mismatched_sizes=True
    )
    
    model.load_state_dict(torch.load('swin.pth', map_location='cpu'))
    model.eval()
    return model, AutoFeatureExtractor.from_pretrained("microsoft/swin-tiny-patch4-window7-224")

model, feature_extractor = load_model()

glasses_recommendations = {
    "Heart": "Rimless (tanpa bingkai bawah)",
    "Oblong": "Kotak",
    "Oval": "Berbagai bentuk bingkai",
    "Round": "Kotak",
    "Square": "Oval atau bundar"
}

def preprocess_image(image):
    results = mp_face_detection.process(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))

    if results.detections:
        detection = results.detections[0]
        bbox = detection.location_data.relative_bounding_box
        h, w, _ = image.shape
        x1 = int(bbox.xmin * w)
        y1 = int(bbox.ymin * h)
        x2 = int((bbox.xmin + bbox.width) * w)
        y2 = int((bbox.ymin + bbox.height) * h)
        image = image[y1:y2, x1:x2]
    else:
        raise ValueError("No face detected")
    
    image = cv2.resize(image, (224, 224))
    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
    
    # Use feature extractor with proper batching
    inputs = feature_extractor(
        images=[image],  # Pass as a list of images
        return_tensors="pt",
        padding=True  # Enable padding as a safety measure
    )
    return inputs['pixel_values']

def predict_face_shape(image):
    image_tensor = preprocess_image(image).to(torch.device("cuda" if torch.cuda.is_available() else "cpu"))
    
    with torch.no_grad():
        outputs = model(image_tensor)
        predicted_class_idx = torch.argmax(outputs.logits, dim=1).item()
    
    return id2label[predicted_class_idx]

# Streamlit UI
st.title("Face Shape & Glasses Recommender")
st.write("Upload a face photo for shape analysis and glasses recommendations")

uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])

if uploaded_file is not None:
    image = Image.open(uploaded_file).convert('RGB')
    img_array = np.array(image)
    
    st.image(image, caption='Uploaded Image', use_column_width=True)
    
    try:
        with st.spinner('Analyzing...'):
            prediction = predict(cv2.cvtColor(img_array, cv2.COLOR_RGB2BGR))
            recommendation = glasses_recommendations[prediction]
            
        st.success(f"Predicted Face Shape: **{prediction}**")
        st.info(f"Recommended Glasses Frame: **{recommendation}**")
    except Exception as e:
        st.error(f"Error: {str(e)}")