# --- Fish Classifier (CNN from scratch) --------------------------------------- import streamlit as st import numpy as np import pandas as pd from PIL import Image import tensorflow as tf from tensorflow.keras.models import load_model # ----------------------------------------------------------------------------- # PAGE CONFIG # ----------------------------------------------------------------------------- st.set_page_config( page_title="Fish Classifier (CNN – From Scratch)", layout="centered" ) # ----------------------------------------------------------------------------- # SETTINGS # ----------------------------------------------------------------------------- MODEL_PATH = "src/best_fish_cnn_new.keras" IMG_SIZE = (224, 224) CLASS_NAMES = [ "Black Sea Sprat", "Gilt-Head Bream", "Horse Mackerel", "Red Mullet", "Red Sea Bream", "Sea Bass", "Shrimp", "Striped Red Mullet", "Trout" ] # ----------------------------------------------------------------------------- # MODEL LOADING (CACHED) # ----------------------------------------------------------------------------- @st.cache_resource def load_model_cached(): model = tf.keras.models.load_model(MODEL_PATH, compile=False) return model model = load_model_cached() # ----------------------------------------------------------------------------- # HELPER FUNCTIONS # ----------------------------------------------------------------------------- def preprocess_image(image: Image.Image) -> np.ndarray: """Resize and normalize image for the CNN model.""" image = image.convert("RGB") image = image.resize(IMG_SIZE) img_array = np.array(image).astype("float32") / 255.0 img_array = np.expand_dims(img_array, axis=0) return img_array def predict_image(image: Image.Image): """Run model prediction and return class label and probability vector.""" img_batch = preprocess_image(image) preds = model.predict(img_batch) probs = preds[0] pred_index = int(np.argmax(probs)) pred_class = CLASS_NAMES[pred_index] return pred_class, probs # ----------------------------------------------------------------------------- # UI – HEADER # ----------------------------------------------------------------------------- st.title("🐟 Fish Species Classifier (CNN From Scratch)") st.write( "Upload a fish image and the model will predict one of **9 fish species**. " "This model is a custom Convolutional Neural Network (**no transfer learning**) " "trained on the Kaggle fish dataset." ) # ----------------------------------------------------------------------------- # SIDEBAR INFO # ----------------------------------------------------------------------------- st.sidebar.header("â„šī¸ About this app") st.sidebar.markdown( """ **Model:** Custom CNN (no transfer learning) **Input size:** 224 × 224 **Classes:** 9 fish species **Validation performance:** - Accuracy: ~84% - Top-3 accuracy: ~95% **Author:** Enes Yamac """ ) # ----------------------------------------------------------------------------- # FILE UPLOAD # ----------------------------------------------------------------------------- uploaded_file = st.file_uploader( "📤 Upload a fish image (JPG/PNG)", type=["jpg", "jpeg", "png"] ) # ----------------------------------------------------------------------------- # PREDICTION + OUTPUT # ----------------------------------------------------------------------------- if uploaded_file is not None: image = Image.open(uploaded_file) st.image(image, caption="Uploaded image", use_container_width=True) with st.spinner("🔍 Predicting..."): pred_class, probs = predict_image(image) st.subheader("🔎 Prediction") st.write(f"**Predicted class:** `{pred_class}`") # Wahrscheinlichkeiten vorbereiten prob_dict = {cls: float(p) for cls, p in zip(CLASS_NAMES, probs)} prob_df = ( pd.DataFrame.from_dict(prob_dict, orient="index", columns=["Probability"]) .sort_values("Probability", ascending=False) ) st.subheader("📊 Class probabilities") st.bar_chart(prob_df) with st.expander("Show numeric probabilities"): st.dataframe(prob_df.style.format("{:.2%}")) else: st.info("Please upload an image to start.")