Create app.py

app.py

@@ -0,0 +1,42 @@
+import gradio as gr
+
+import tensorflow as tf
+import tensorflow_hub as hub
+
+ckpt_type = '1k'
+tf_hub_url = 'gs://cloud-tpu-checkpoints/efficientnet/v2/hub/efficientnetv2-s/classification'
+
+m = hub.KerasLayer(tf_hub_url, trainable=False)
+m.build([None, 224, 224, 3])  # Batch input shape.
+
+def get_imagenet_labels(filename):
+  labels = []
+  with open(filename, 'r') as f:
+    for line in f:
+      labels.append(line.split('\t')[1][:-1])  # split and remove line break.
+  return labels
+  
+classes = get_imagenet_labels("imagenet1k_labels.txt")
+
+def classify(image):
+  image = tf.keras.preprocessing.image.img_to_array(image)
+  image = (image - 128.) / 128.
+  logits = m(tf.expand_dims(image, 0), False)
+  pred = tf.keras.layers.Softmax()(logits)
+  idx = tf.argsort(logits[0])[::-1][0].numpy()
+  return classes[idx]
+  
+title = "Interactive demo: EfficientNetV2"
+description = "Demo for Google's EfficientNetV2. EfficientNetV2 (accepted at ICML 2021) consists of convolutional neural networks that aim for fast training speed for relatively small-scale datasets, such as ImageNet1k."
+article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2104.00298'>EfficientNetV2: Smaller Models and Faster Training</a> | <a href='https://github.com/google/automl/tree/master/efficientnetv2'>Github Repo</a> | <a href='https://ai.googleblog.com/2021/09/toward-fast-and-accurate-neural.html'>Blog Post</a></p>"
+
+iface = gr.Interface(fn=classify, 
+  inputs=gr.inputs.Image(label="image"),
+  outputs='text',
+  title=title,
+  description=description,
+  enable_queue=True,
+  examples=[['panda.jpeg'], ["llamas.jpeg"], ["hot dog.png"]],
+  article=article)
+
+iface.launch()