app.py

import numpy as np
import pandas as pd
import cv2
import torch
from tensorflow import keras
from PIL import Image
from transformers import ViTForImageClassification, ViTFeatureExtractor
from tensorflow.keras.models import load_model
import streamlit as st

st.title("Skin Cancer Classification App")

# Load TensorFlow models
models = {
    "Le_Net": load_model('LeNet_5.h5'),
    "Simple_CNN": load_model('Simple CNN.h5'),
    "Alex_Net": load_model('AlexNet.h5'),
    "Deeper_CNN": load_model('Deeper CNN.h5'),
}

# Load PyTorch ViT model
vit_model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224', num_labels=7)
vit_model.load_state_dict(torch.load('./vit_skin_cancer_model.pth'))
vit_model.eval()  # Set the model to evaluation mode

# Add the PyTorch model to the models dictionary
models["ViT_Model"] = vit_model

# Allow user to select model
model_name = st.selectbox("Choose a model", list(models.keys()))
model = models[model_name]

# Upload Image
file = st.file_uploader("Upload an image", type=["png", "jpg", "jpeg"])

true_file = pd.read_csv("HAM10000_metadata.csv")

classes = {
    4: ('nv', 'melanocytic nevi'), 
    6: ('mel', 'melanoma'),
    2: ('bkl', 'benign keratosis-like lesions'), 
    1: ('bcc', 'basal cell carcinoma'),
    5: ('vasc', 'pyogenic granulomas and hemorrhage'),
    0: ('akiec', 'Actinic keratoses and intraepithelial carcinomae'),
    3: ('df', 'dermatofibroma')
}

classes_map = {
    'nv': 'melanocytic nevi', 
    'mel': 'melanoma',
    'bkl': 'benign keratosis-like lesions',
    'bcc': 'basal cell carcinoma',
    'vasc': 'pyogenic granulomas and hemorrhage',
    'akiec': 'Actinic keratoses and intraepithelial carcinomae',
    'df': 'dermatofibroma'
}

if file is not None:
    file_bytes = np.asarray(bytearray(file.read()), dtype=np.uint8)
    opencv_image = cv2.imdecode(file_bytes, 1)
    
    # Resize image for TensorFlow models
    img1 = cv2.resize(opencv_image, (32, 32))
    
    if model_name == "ViT_Model":
        # PyTorch model inference
        feature_extractor = ViTFeatureExtractor.from_pretrained('google/vit-base-patch16-224')
        image = feature_extractor(images=opencv_image, return_tensors="pt")['pixel_values']
        with torch.no_grad():
            outputs = model(image)
            class_ind = outputs.logits.argmax(-1).item()
            class_name = classes[class_ind]
    else:
        # TensorFlow model inference
        result = model.predict(img1.reshape(1, 32, 32, 3))
        max_prob = max(result[0])
        class_ind = list(result[0]).index(max_prob)
        class_name = classes[class_ind]

    # Display image and result 
    col1, col2 = st.columns(2)
    with col1:
        st.header("Input Image")
        st.image(opencv_image, channels="BGR")
    with col2:
        st.header("Results")
        
        if file:
            name = file.name.split(".")[0]
        if name in true_file['image_id'].values: 
            st.write("True Label: ", classes_map[true_file.loc[true_file['image_id']==name, 'dx'].iloc[0]])
            st.write("Prediction:", class_name[1])
        else:
            st.write("No match")