sync with remote

app.py

@@ -0,0 +1,143 @@
+import streamlit as st
+import tensorflow as tf
+import os
+import requests
+import tempfile
+import matplotlib.pyplot as plt
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Flatten, Dense, Reshape
+from tensorflow.keras.losses import SparseCategoricalCrossentropy
+from io import StringIO
+
+# Constants for dataset information
+TRAIN_FILE = "train_images.tfrecords"
+VAL_FILE = "val_images.tfrecords"
+TRAIN_URL = "https://huggingface.co/datasets/louiecerv/cardiac_images/resolve/main/train_images.tfrecords"
+VAL_URL = "https://huggingface.co/datasets/louiecerv/cardiac_images/resolve/main/val_images.tfrecords"
+
+# Use a persistent temp directory
+tmpdir = tempfile.gettempdir()
+
+# Function to download a file with progress display
+def download_file(url, local_filename, target_dir):
+    os.makedirs(target_dir, exist_ok=True)
+    filepath = os.path.join(target_dir, local_filename)
+
+    if os.path.exists(filepath):
+        st.write(f"File already exists: {filepath}")
+        return filepath
+
+    with requests.get(url, stream=True) as r:
+        r.raise_for_status()
+        total_size = int(r.headers.get('content-length', 0))
+
+        progress_bar = st.empty()  # Create a placeholder
+
+        with open(filepath, 'wb') as f:
+            downloaded_size = 0
+            for chunk in r.iter_content(chunk_size=8192):
+                if chunk:
+                    f.write(chunk)
+                    downloaded_size += len(chunk)
+                    progress_percent = int(downloaded_size / total_size * 100)
+                    progress_bar.progress(progress_percent, text=f"Downloading {local_filename}...")
+
+    return filepath
+
+# Download only if files are missing
+train_file_path = download_file(TRAIN_URL, TRAIN_FILE, tmpdir)
+val_file_path = download_file(VAL_URL, VAL_FILE, tmpdir)
+
+# Dictionary describing the fields stored in TFRecord
+image_feature_description = {
+    'height': tf.io.FixedLenFeature([], tf.int64),
+    'width': tf.io.FixedLenFeature([], tf.int64),
+    'depth': tf.io.FixedLenFeature([], tf.int64),
+    'name': tf.io.FixedLenFeature([], tf.string),
+    'image_raw': tf.io.FixedLenFeature([], tf.string),
+    'label_raw': tf.io.FixedLenFeature([], tf.string),
+}
+
+# Helper function to parse the image and label data from TFRecord
+def _parse_image_function(example_proto):
+    return tf.io.parse_single_example(example_proto, image_feature_description)
+
+# Function to read and decode an example from the dataset
+@tf.function
+def read_and_decode(example):
+    image_raw = tf.io.decode_raw(example['image_raw'], tf.int64)
+    image_raw.set_shape([65536])
+    image = tf.reshape(image_raw, [256, 256, 1])
+
+    image = tf.cast(image, tf.float32) * (1. / 1024)
+
+    label_raw = tf.io.decode_raw(example['label_raw'], tf.uint8)
+    label_raw.set_shape([65536])
+    label = tf.reshape(label_raw, [256, 256, 1])
+
+    return image, label
+
+# Load and parse datasets
+raw_training_dataset = tf.data.TFRecordDataset(train_file_path)
+raw_val_dataset = tf.data.TFRecordDataset(val_file_path)
+
+parsed_training_dataset = raw_training_dataset.map(_parse_image_function)
+parsed_val_dataset = raw_val_dataset.map(_parse_image_function)
+
+# Prepare datasets
+tf_autotune = tf.data.experimental.AUTOTUNE
+train = parsed_training_dataset.map(read_and_decode, num_parallel_calls=tf_autotune)
+val = parsed_val_dataset.map(read_and_decode)
+
+BUFFER_SIZE = 10
+BATCH_SIZE = 1
+
+train_dataset = train.cache().shuffle(BUFFER_SIZE).batch(BATCH_SIZE).repeat()
+train_dataset = train_dataset.prefetch(buffer_size=tf_autotune)
+test_dataset = val.batch(BATCH_SIZE)
+
+st.write(train_dataset)
+
+def display(display_list):
+    fig = plt.figure(figsize=(10, 10))
+    title = ['Input Image', 'Label']
+
+    for i in range(len(display_list)):
+        ax = fig.add_subplot(1, len(display_list), i + 1)
+        display_resized = tf.reshape(display_list[i], [256, 256])
+        ax.set_title(title[i])
+        ax.imshow(display_resized, cmap='gray')
+        ax.axis('off')
+
+    st.pyplot(fig)
+
+# Streamlit app interface
+st.title("Cardiac Images Dataset")
+
+# Display sample images
+for image, label in train.take(2):
+    sample_image, sample_label = image, label
+    display([sample_image, sample_label])
+
+tf.keras.backend.clear_session()
+
+# set up the model architecture
+model = tf.keras.models.Sequential([
+    Flatten(input_shape=[256, 256, 1]),
+    Dense(64, activation='relu'),
+    Dense(256*256*2, activation='softmax'),
+    Reshape((256, 256, 2))
+])
+
+# specify how to train the model with algorithm, the loss function and metrics
+model.compile(
+    optimizer='adam',
+    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
+    metrics=['accuracy'])
+
+# Capture the model summary
+model_summary = StringIO()
+model.summary(print_fn=lambda x: model_summary.write(x + '\n'))
+
+# Display the model summary in Streamlit
+st.markdown(model_summary.getvalue())

requirements.txt

@@ -0,0 +1,5 @@
+streamlit
+datasets
+tensorflow
+pandas
+matplotlib