uploading project

Author: lucasbazante

.ipynb_checkpoints/image_similarity-checkpoint.ipynb

@@ -0,0 +1,85 @@
+import tensorflow_datasets as tfds
+
+import numpy as np
+import matplotlib.pyplot as plt
+
+import tensorflow as tf
+from tensorflow import keras
+from keras import layers
+
+import functools
+
+from sklearn.neighbors import NearestNeighbors # similarity
+
+import ipywidgets as w # interface
+
+
+ds = tfds.load("imagenette/160px")
+
+
+def extract_image(example):
+    image = example['image']
+    return image
+
+def preprocess_image(image, height, width):
+    image = tf.image.resize(image, (height, width))
+    image = tf.cast(image, tf.float32) / 255.0
+    return image
+
+def get_image_batches(batch_size = 128, height = 256, width = 256):
+    partial_preprocess_image = functools.partial(preprocess_image, height=height, width=width)
+    train_ds = ds['train']
+    train_ds = ( train_ds.map(extract_image)
+                .map(partial_preprocess_image)
+                .cache()
+                .shuffle(buffer_size=1000)
+                .batch(batch_size)
+                .prefetch(tf.data.AUTOTUNE)
+                )
+    return train_ds
+
+
+BATCH_SIZE = 32
+IMG_WIDTH = IMG_HEIGHT = 32
+train_ds = get_image_batches(batch_size = BATCH_SIZE, height = IMG_HEIGHT, width = IMG_WIDTH)
+
+
+images = np.array([img for batch in train_ds.take(300) for img in batch])
+print(images.shape)
+
+
+encoder = keras.models.load_model("saved_models/encoder.keras")
+encoder.summary()
+
+
+features = encoder.predict(images, batch_size = BATCH_SIZE)
+
+knn = NearestNeighbors(n_neighbors=5, metric="cosine")
+knn.fit(features)
+
+
+def show_similar_images(start_image_idx, n_neighbors=10):
+    input_image = images[start_image_idx]
+    input_image = np.expand_dims(input_image, 0)
+    fig, axes = plt.subplots(1, n_neighbors+1, figsize=(n_neighbors * 2, 5))
+
+    feature = encoder.predict(input_image)
+    distances, nbors = knn.kneighbors(feature, n_neighbors=n_neighbors)
+    distances, nbors = distances[0], nbors[0]
+    nbor_images = [images[i] for i in nbors]
+
+    for i in range(len(nbor_images)+1):
+        ax = axes[i]
+        ax.set_axis_off()
+        if i == 0:
+            ax.imshow(input_image.squeeze(0))
+            ax.set_title("Input")
+        else:
+            ax.imshow(nbor_images[i-1])
+            ax.set_title(f"Sim: {1 - distances[i-1]:.2f}")
+    plt.show()
+
+w.interact(show_similar_images, 
+    start_image_idx=w.IntSlider(max=len(images)-1, continuous_update=False),
+    n_neighbors=w.IntSlider(min=2, value=5, max=10, continuous_update=False),
+)