from fastapi import FastAPI import pickle import uvicorn import pandas as pd import shutil import cv2 import mediapipe as mp from werkzeug.utils import secure_filename import tensorflow as tf import os from flask import Flask, jsonify, request, flash, redirect, url_for from pyngrok import ngrok from fastapi import FastAPI, HTTPException, File, UploadFile, Request from fastapi.staticfiles import StaticFiles from fastapi.responses import JSONResponse from pydantic import BaseModel import subprocess from file_processing import FileProcess from get_load_data import GetLoadData from data_preprocess import DataProcessing from train_pred import TrainPred app = FastAPI() public_url = "https://lambang0902-test-space.hf.space" app.mount("/static", StaticFiles(directory="static"), name="static") # Tempat deklarasi variabel-variabel penting filepath = "" list_class = ['Diamond','Oblong','Oval','Round','Square','Triangle'] list_folder = ['Training', 'Testing'] face_crop_img = True face_landmark_img = True landmark_extraction_img = True # ----------------------------------------------------- # ----------------------------------------------------- # Tempat deklarasi model dan sejenisnya selected_model = tf.keras.models.load_model(f'models/fc_model_1.h5', compile=False) face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_alt2.xml') mp_drawing = mp.solutions.drawing_utils mp_face_mesh = mp.solutions.face_mesh drawing_spec = mp_drawing.DrawingSpec(thickness=1, circle_radius=1) # ----------------------------------------------------- # ----------------------------------------------------- # Tempat setting server UPLOAD_FOLDER = './upload' UPLOAD_MODEL = './models' ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg','zip','h5'} # ----------------------------------------------------- #Endpoints #Root endpoints @app.get("/") async def root(): # Dapatkan URL publik dari ngrok ngrok_url = "Tidak Ada URL Publik (ngrok belum selesai memulai)" return {"message": "Hello, World!", "ngrok_url": ngrok_url} # #----------------------------------------------------- # data_processor = DataProcessing() data_train_pred = TrainPred() # import random def preprocessing(filepath): folder_path = './static/temporary' shutil.rmtree(folder_path) os.mkdir(folder_path) # data_processor.detect_landmark(data_processor.face_cropping_pred(filepath)) data_processor.enhance_contrast_histeq(data_processor.face_cropping_pred(filepath)) files = os.listdir(folder_path) index = 0 for file_name in files: file_ext = os.path.splitext(file_name)[1] new_file_name = str(index) + "_" + str(random.randint(1, 100000)) + file_ext os.rename(os.path.join(folder_path, file_name), os.path.join(folder_path, new_file_name)) index += 1 print("Tungu sampai selesaiii") train_datagen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1 / 255.) test_datagen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1 / 255.) ## ------------------------------------------------------------------------- ## API UNTUK MELAKUKAN PROSES PREDIKSI ## ------------------------------------------------------------------------- @app.post('/upload/file',tags=["Predicting"]) async def upload_file(picture: UploadFile): file_extension = picture.filename.split('.')[-1].lower() if file_extension not in ALLOWED_EXTENSIONS: raise HTTPException(status_code=400, detail='Invalid file extension') os.makedirs(UPLOAD_FOLDER, exist_ok=True) file_path = os.path.join(UPLOAD_FOLDER, secure_filename(picture.filename)) with open(file_path, 'wb') as f: f.write(picture.file.read()) try: processed_img = preprocessing(cv2.imread(file_path)) except Exception as e: os.remove(file_path) raise HTTPException(status_code=500, detail=f'Error processing image: {str(e)}') return JSONResponse(content={'message': 'File successfully uploaded'}, status_code=200) @app.get('/get_images', tags=["Predicting"]) def get_images(): folder_path = "./static/temporary" files = [f for f in os.listdir(folder_path) if os.path.isfile(os.path.join(folder_path, f))] urls = [] for i in range(0, 3): url = f'{public_url}/static/temporary/{files[i]}' urls.append(url) bentuk, persentase = data_train_pred.prediction(selected_model) return {'urls': urls, 'bentuk_wajah':bentuk[0], 'persen':persentase} ## ------------------------------------------------------------------------- ## API UNTUK MELAKUKAN PROSES TRAINING ## ------------------------------------------------------------------------- # Model pydantic untuk validasi body class TrainingParams(BaseModel): optimizer: str epoch: int batchSize: int @app.post('/upload/dataset', tags=["Training"]) async def upload_data(dataset: UploadFile): if dataset.filename == '': raise HTTPException(status_code=400, detail='No file selected for uploading') # Buat path lengkap untuk menyimpan file file_path = os.path.join(UPLOAD_FOLDER, dataset.filename) # Simpan file ke folder yang ditentukan with open(file_path, "wb") as file_object: file_object.write(dataset.file.read()) # Panggil fungsi untuk mengekstrak file jika perlu FileProcess.extract_zip(file_path) return {'message': 'File successfully uploaded'} @app.post('/set_params', tags=["Training"]) async def set_params(request: Request, params: TrainingParams): global optimizer, epoch, batch_size optimizer = params.optimizer epoch = params.epoch batch_size = params.batchSize response = {'message': 'Set parameter sukses'} return response @app.get('/get_info_data', tags=["Training"]) def get_info_prepro(): global optimizer, epoch, batch_size training_counts = GetLoadData.get_training_file_counts().json testing_counts = GetLoadData.get_testing_file_counts().json response = { "optimizer": optimizer, "epoch": epoch, "batch_size": batch_size, "training_counts": training_counts, "testing_counts": testing_counts } return response @app.get('/get_images_preprocess', tags=["Training"]) def get_random_images_crop(): images_face_landmark = GetLoadData.get_random_images(tahap="Face Landmark",public_url=public_url) images_face_extraction = GetLoadData.get_random_images(tahap="landmark Extraction", public_url=public_url) response = { "face_landmark": images_face_landmark, "landmark_extraction": images_face_extraction } return response @app.get('/do_preprocessing', tags=["Training"]) async def do_preprocessing(): try: data_train_pred.do_pre1(test="") data_train_pred.do_pre2(test="") return {'message': 'Preprocessing sukses'} except Exception as e: # Tangani kesalahan dan kembalikan respons kesalahan error_message = f'Error during preprocessing: {str(e)}' raise HTTPException(status_code=500, detail=error_message) @app.get('/do_training', tags=["Training"]) def do_training(): global epoch folder = "" if (face_landmark_img == True and landmark_extraction_img == True): folder = "Landmark Extraction" elif (face_landmark_img == True and landmark_extraction_img == False): folder = "Face Landmark" # -------------------------------------------------------------- train_dataset_path = f"./static/dataset/{folder}/Training/" test_dataset_path = f"./static/dataset/{folder}/Testing/" train_image_df, test_image_df = GetLoadData.load_image_dataset(train_dataset_path, test_dataset_path) train_gen, test_gen = data_train_pred.data_configuration(train_image_df, test_image_df) model = data_train_pred.model_architecture() result = data_train_pred.train_model(model, train_gen, test_gen, epoch) # Mengambil nilai akurasi training dan validation dari objek result train_acc = result.history['accuracy'][-1] val_acc = result.history['val_accuracy'][-1] # Plot accuracy data_train_pred.plot_accuracy(result=result, epoch=epoch) acc_url = f'{public_url}/static/accuracy_plot.png' # Plot loss data_train_pred.plot_loss(result=result, epoch=epoch) loss_url = f'{public_url}/static/loss_plot.png' # Confusion Matrix data_train_pred.plot_confusion_matrix(model, test_gen) conf_url = f'{public_url}/static/confusion_matrix.png' return jsonify({'train_acc': train_acc, 'val_acc': val_acc, 'plot_acc': acc_url, 'plot_loss':loss_url,'conf':conf_url})