File size: 4,339 Bytes
425b51a
 
818c12f
 
425b51a
 
 
 
 
571b79d
 
425b51a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
571b79d
425b51a
 
571b79d
425b51a
 
 
 
 
 
 
 
 
 
 
 
 
571b79d
425b51a
 
 
571b79d
 
 
425b51a
571b79d
 
 
 
 
 
 
 
 
 
 
 
425b51a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
571b79d
 
 
 
 
 
 
 
 
 
425b51a
 
 
 
 
 
 
 
 
 
 
 
571b79d
425b51a
 
818c12f
425b51a
571b79d
425b51a
 
 
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
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
# app_rice_resnet50.py
import streamlit as st
import numpy as np
import pandas as pd
from PIL import Image
import tensorflow as tf
from tensorflow.keras.applications.resnet50 import preprocess_input

# ----------------- SETTINGS -----------------
# Falls das Modell im Root liegt, nimm: "./best_rice_resnet50.keras"
MODEL_PATH = "src/best_rice_resnet50.keras"
IMG_SIZE = (224, 224)

CLASS_NAMES = [
    "Arborio",
    "Basmati",
    "Ipsala",
    "Jasmine",
    "Karacadag",
]

# ----------------- MODEL LOADING -----------------
@st.cache_resource
def load_resnet_model():
    try:
        model = tf.keras.models.load_model(MODEL_PATH, compile=False)
        return model
    except Exception as e:
        st.error(f"Error loading model: {e}")
        return None


def preprocess_image(image: Image.Image) -> np.ndarray:
    """Resize image and apply ResNet50 preprocessing."""
    img = image.convert("RGB")
    img = img.resize(IMG_SIZE)
    img_array = np.array(img).astype("float32")
    img_array = preprocess_input(img_array)
    img_array = np.expand_dims(img_array, axis=0)
    return img_array


def predict_image(model, image_array: np.ndarray):
    """Safe prediction with Top-3; robust gegen Längen-Mismatch."""
    preds = model.predict(image_array, verbose=0)[0]
    probs = tf.nn.softmax(preds).numpy()

    n_model = len(probs)
    n_labels = len(CLASS_NAMES)
    n = min(n_model, n_labels)

    # Hauptklasse (nur im Bereich [0, n))
    main_idx = int(np.argmax(probs[:n]))
    pred_class = CLASS_NAMES[main_idx]
    pred_conf = float(probs[main_idx])

    # Top-k (hier 3) nur im gültigen Bereich
    k = min(3, n)
    topk_idx = np.argsort(probs[:n])[-k:][::-1]
    topk = [(CLASS_NAMES[i], float(probs[i])) for i in topk_idx]

    # Nur die ersten n Wahrscheinlichkeiten zurückgeben
    return pred_class, pred_conf, probs[:n], topk, n_model, n_labels


# ----------------- PAGE CONFIG -----------------
st.set_page_config(
    page_title="Rice Classifier – ResNet50 (Transfer Learning)",
    page_icon="🌾",
    layout="wide",
)

# ----------------- SIDEBAR -----------------
st.sidebar.title("🌾 Rice ResNet50 Classifier")
st.sidebar.write(
    "This app uses a ResNet50 transfer learning model to classify rice grains."
)

st.sidebar.subheader("Rice classes")
for idx, name in enumerate(CLASS_NAMES):
    st.sidebar.write(f"- **{idx}** → {name}")

st.sidebar.markdown("---")
st.sidebar.caption("Author: Enes Yamac")

# ----------------- MAIN PAGE -----------------
st.title("🌾 Rice Classifier – ResNet50 (Transfer Learning)")
st.write("Upload a rice grain image. The model will predict the rice type and show top-3 results.")

uploaded_file = st.file_uploader(
    "Upload a rice image (JPG or PNG)",
    type=["jpg", "jpeg", "png"],
)

if uploaded_file is not None:
    image = Image.open(uploaded_file)

    col1, col2 = st.columns([1, 1])

    with col1:
        st.subheader("Uploaded image")
        st.image(image, width=500)

    with st.spinner("Loading ResNet50 model and making prediction..."):
        model = load_resnet_model()

    if model is not None:
        img_array = preprocess_image(image)
        pred_class, pred_conf, probs, top3, n_model, n_labels = predict_image(
            model, img_array
        )

        # kurze Warnung, falls Modell/Labels nicht exakt matchen
        if n_model != n_labels:
            st.warning(
                f"Model outputs {n_model} classes, CLASS_NAMES has {n_labels}. "
                f"Only the first {min(n_model, n_labels)} are shown."
            )

        with col2:
            st.subheader("Prediction")
            st.markdown(f"**Predicted class:** `{pred_class}`")
            st.markdown(f"**Confidence:** `{pred_conf:.2%}`")

            st.subheader("Top-3 predictions")
            for name, p in top3:
                st.write(f"- {name}: {p:.2%}")

            st.subheader("Class probabilities")
            df = pd.DataFrame(
                {"class": CLASS_NAMES[: len(probs)], "probability": probs}
            ).set_index("class")
            st.bar_chart(df)

            with st.expander("Show probabilities as text"):
                for name, p in zip(CLASS_NAMES[: len(probs)], probs):
                    st.write(f"- {name}: {p:.2%}")
else:
    st.info("Please upload a rice image to get a prediction.")