Zeeshan01/hyper

1import gradio as gr2import tensorflow as tf3import numpy as np4from PIL import Image5 6 7 8 9 10 11# Initial parameters for pretrained model12IMG_SIZE = 30013 14 15 16labelInfo = {17    'lower-gi-tract anatomical-landmarks cecum': 0,18    'lower-gi-tract anatomical-landmarks ileum': 1,19    'lower-gi-tract anatomical-landmarks retroflex-rectum': 2,20    'lower-gi-tract pathological-findings hemorrhoids': 3,21    'lower-gi-tract pathological-findings polyps': 4,22    'lower-gi-tract pathological-findings ulcerative-colitis-grade-0-1': 5,23    'lower-gi-tract pathological-findings ulcerative-colitis-grade-1': 6,24    'lower-gi-tract pathological-findings ulcerative-colitis-grade-1-2': 7,25    'lower-gi-tract pathological-findings ulcerative-colitis-grade-2': 8,26    'lower-gi-tract pathological-findings ulcerative-colitis-grade-2-3': 9,27    'lower-gi-tract pathological-findings ulcerative-colitis-grade-3': 10,28    'lower-gi-tract quality-of-mucosal-views bbps-0-1': 11,29    'lower-gi-tract quality-of-mucosal-views bbps-2-3': 12,30    'lower-gi-tract quality-of-mucosal-views impacted-stool': 13,31    'lower-gi-tract therapeutic-interventions dyed-lifted-polyps': 14,32    'lower-gi-tract therapeutic-interventions dyed-resection-margins': 15,33    'upper-gi-tract anatomical-landmarks pylorus': 16,34    'upper-gi-tract anatomical-landmarks retroflex-stomach': 17,35    'upper-gi-tract anatomical-landmarks z-line': 18,36    'upper-gi-tract pathological-findings barretts': 19,37    'upper-gi-tract pathological-findings barretts-short-segment': 20,38    'upper-gi-tract pathological-findings esophagitis-a': 21,39    'upper-gi-tract pathological-findings esophagitis-b-d': 2240}41# Load the model from the H5 file42model = tf.keras.models.load_model('model/Hyper.h5')43 44# Define the prediction function45def predict(img):46    img_height = 30047    img_width = 30048 49    # Convert the NumPy array to a PIL Image object50    pil_img = Image.fromarray(img)51 52    # Resize the image using the PIL Image object53    pil_img = pil_img.resize((img_height, img_width))54 55    # Convert the PIL Image object to a NumPy array56    x = tf.keras.preprocessing.image.img_to_array(pil_img)57 58    x = x.reshape(1, img_height, img_width, 3)59    np.set_printoptions(formatter={'float': '{: 0.3f}'.format})60 61 62    predi = model.predict(x)63    accuracy_of_class = '{:.1f}'.format(predi[0][np.argmax(predi)] * 100) + "%"64    classes = list(labelInfo.keys())[np.argmax(predi)]65    context = {66        'predictedLabel': classes,67        # 'y_class': y_class,68        # 'z_class': z_class,69        'accuracy_of_class': accuracy_of_class70    }71 72 73   74    return context75 76 77 78demo = gr.Interface(fn=predict, inputs="image", outputs="text" , examples=[["O1.jpg"],["O2.jpg"],["O3.jpg"]],)79 80demo.launch()81 82