import tensorflow as tf import numpy as np from PIL import Image import os # from matplotlib import image as mpimg # from matplotlib import pyplot as plt class api(): height=64 width=64 channels=3 model_name = 'cnn_model' classes = { 0 : 'Zero' , 1 : 'One' , 2 : 'Two' , 3 : 'Three' , 4 : 'Four' , 5 : 'Five' } def reset_graph(self,seed=42): tf.reset_default_graph() tf.set_random_seed(seed) np.random.seed(seed) def __init__(self,upload_path='uploads'): self.upload_path = upload_path # self.model_name = 'cnn_model' print('print',os.path.join('signs_api','{}.meta'.format(self.model_name))) # self.import_meta = tf.train.import_meta_graph(os.path.join('signs_api','{}.meta'.format(self.model_name))) def predict(self,im): try : # im = Image.open( os.path.join(self.upload_path,filename) ) #image size size=(self.height,self.width) #resize image out = im.resize(size) test_image = np.array(out.getdata()) test_image = test_image.reshape((-1,self.height,self.width,self.channels)) # to make this notebook's output stable across runs self.reset_graph() # import meta from directory # import_meta = tf.train.import_meta_graph('{}.meta'.format(self.model_name)) import_meta = tf.train.import_meta_graph(os.path.join('signs_api','{}.meta'.format(self.model_name))) with tf.Session() as sess: # tf.train.latest_checkpoint() also works import_meta.restore(sess,'{}.ckpt'.format( os.path.join('signs_api',self.model_name) ) ) # W1_val = sess.graph.get_tensor_by_name('W1:0') # X_val = sess.graph.get_tensor_by_name('Placeholder:0') ArgMax = sess.graph.get_tensor_by_name('ArgMax:0') ArgMax_val = ArgMax.eval({ 'Placeholder:0' : test_image }) # graph = tf.get_default_graph() # for op in graph.get_operations(): # print(op.name) # print('W1_val',W1_val) # print('X_val',X_val) print('ArgMax',ArgMax_val) index = ArgMax_val.tolist()[0] class_val = self.classes[index] # os.remove(os.path.join(self.upload_path,filename)) return { 'value' : index , 'class' : class_val } except (OSError,IOError) as e: print('error',e) return { 'error' : True }