## THIS IS FOR DENOISING import gradio as gr from PIL import Image import tensorflow as tf import tensorflow_hub as hub import numpy as np import requests import cv2 def resize_image(image, target_dim): shape = tf.cast(tf.shape(image)[1:-1], tf.float32) short_dim = min(shape) scale = target_dim / short_dim new_shape = tf.cast(shape * scale, tf.int32) image = tf.image.resize(image, new_shape) image = tf.image.resize_with_crop_or_pad(image, target_dim, target_dim) return image def process_image(image_path, target_dim=256): input_img = cv2.cvtColor(image_path, cv2.COLOR_BGR2RGB) input_img=np.asarray(input_img,np.float32)/255.0 input_img = tf.expand_dims(input_img, axis=0) input_img = resize_image(input_img, target_dim) return input_img def infer(image_path: str, model: tf.keras.Model, input_resolution=(256, 256)): preprocessed_image = process_image(image_path, input_resolution[0]) preds = model.predict(preprocessed_image) if isinstance(preds, list): preds = preds[-1] if isinstance(preds, list): preds = preds[-1] preds = np.array(preds[0], np.float32) final_pred_image = np.array((np.clip(preds, 0.0, 1.0)).astype(np.float32)) return final_pred_image input_resolution = (256,256) def get_model(model_url: str, input_resolution: tuple) -> tf.keras.Model: inputs = tf.keras.Input((*input_resolution, 3)) hub_module = hub.KerasLayer(model_url) outputs = hub_module(inputs) return tf.keras.Model(inputs, outputs) denoise_model = get_model('./denoise_graph', input_resolution) def predict_image(image): return infer(image,denoise_model,input_resolution) image=gr.inputs.Image() irface=gr.Interface(fn=predict_image, inputs=image, outputs=image,interpretation='default') irface.launch()