import numpy as np # linear algebra import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv) import tensorflow as tf from PIL import Image import os classes = [ "cocci" , "healthy" , "ncd" , "salmo" , ] def predict(): # Assuming model, classes, and test are already defined result = [] test= load_test_data() for index, i in enumerate(test.filename): img = Image.open(i).convert('RGB') img = img.resize((300, 300 * img.size[1] // img.size[0]), Image.Resampling.LANCZOS) inp_numpy = np.array(img)[None] inp = tf.constant(inp_numpy, dtype='float32') # Get model predictions model = tf.saved_model.load(get_model_path()) class_scores = model(inp)[0].numpy() predicted_class_index = class_scores.argmax() predicted_class = classes[predicted_class_index] confidence = class_scores[predicted_class_index] result.append({"label":predicted_class, "accuracy":confidence}) return result def get_pwd(): # Define the path relative to the script's directory return os.path.dirname(os.path.abspath(__file__)) def get_upload_dir(relative_path='../uploads'): upload_dir = os.path.join(get_pwd(),relative_path) # Get the absolute path to the uploads directory return os.path.abspath(upload_dir) def get_model_path(): current_dir = get_pwd() return os.path.join(current_dir, 'EfficientNet-model') def load_test_data(): Id = [] upload_dir = get_upload_dir() for dirname, _, filenames in os.walk(upload_dir) : for filename in filenames: Id.append(os.path.join(dirname, filename)) test=pd.DataFrame() test=test.assign(filename=Id) test.head() return test