initial commit

README.md

@@ -1,10 +1,10 @@
 ---
 title: FashionMNIST
-emoji: 🐨
+emoji: 📷
 colorFrom: gray
 colorTo: yellow
 sdk: gradio
-sdk_version: 5.1.0
+sdk_version: 4.12.0
 app_file: app.py
 pinned: false
 license: mit

app.py

@@ -0,0 +1,122 @@
+import os
+import random
+
+import cv2
+import gradio as gr
+import joblib
+import numpy as np
+import torch
+
+from models.cnn import Classifier
+from models.feed_forward import FeedForwardClassifier
+
+simple_classifier = Classifier()
+CNN_PATH = "models/classifier_cnn.pth"
+simple_classifier.load_state_dict(torch.load(CNN_PATH))
+
+feed_forward_classifier = FeedForwardClassifier(784)
+FF_PATH = "models/classifier.pth"
+feed_forward_classifier.load_state_dict(torch.load(FF_PATH))
+
+# Required for the classifier
+from sklearn.ensemble import RandomForestClassifier
+
+RF_PATH = "models/fashion_mnist_rf_model.pkl"
+rf_clf = joblib.load(RF_PATH)
+
+
+# Required for the classifier
+from sklearn.svm import SVC
+
+SVM_PATH = "models/fashion_mnist_svm_model.pkl"
+svm_clf = joblib.load(SVM_PATH)
+SVM_PATH_RBF = "models/fashion_mnist_svm_model_rbf.pkl"
+svm_clf_rbf = joblib.load(SVM_PATH_RBF)
+
+
+LABELS = [
+    "T-shirt/top",
+    "Trouser",
+    "Pullover",
+    "Dress",
+    "Coat",
+    "Sandal",
+    "Shirt",
+    "Sneaker",
+    "Bag",
+    "Ankle boot",
+]
+
+
+def classify(img: str):
+    # read image
+    img = cv2.imread(img, cv2.IMREAD_GRAYSCALE)
+    img = cv2.resize(img, (28, 28))
+    img = img / 255.0
+    img = np.array(img).reshape(-1, 1, 28, 28)
+    img = torch.from_numpy(img).float()
+
+    cnn_output = simple_classifier(img)
+    cnn_output = torch.nn.functional.softmax(cnn_output, dim=1)
+    cnn_output_pred = torch.argmax(cnn_output)
+    cnn_output_label = LABELS[cnn_output_pred.item()]
+    cnn_output_str = f"Simple CNN: {cnn_output_label} with {round(torch.max(cnn_output).item()*100, 2)}% confidence"
+
+    feed_forward_output = feed_forward_classifier(img)
+    feed_forward_output = torch.nn.functional.softmax(feed_forward_output, dim=1)
+    feed_forward_output_pred = torch.argmax(feed_forward_output)
+    feed_forward_output_label = LABELS[feed_forward_output_pred.item()]
+    feed_forward_output_str = f"Feed Forward: {feed_forward_output_label} with {round(torch.max(feed_forward_output).item()*100, 2)}% confidence"
+
+    rf_output = rf_clf.predict(img.reshape(1, -1))
+    rf_output_label = LABELS[rf_output[0]]
+    rf_output_str = f"Random Forest: {rf_output_label}"
+
+    svm_output = svm_clf.predict(img.reshape(1, -1))
+    svm_output_label = LABELS[svm_output[0]]
+    svm_output_str = f"SVM with linear kernel: {svm_output_label}"
+
+    svm_output_rbf = svm_clf_rbf.predict(img.reshape(1, -1))
+    svm_output_label_rbf = LABELS[svm_output_rbf[0]]
+    svm_output_str_rbf = f"SVM with RBF kernel: {svm_output_label_rbf}"
+
+    output = (
+        cnn_output_str
+        + "\n"
+        + feed_forward_output_str
+        + "\n"
+        + rf_output_str
+        + "\n"
+        + svm_output_str
+        + "\n"
+        + svm_output_str_rbf
+    )
+    return output
+
+
+folder = "./images"
+examples = []
+for filename in os.listdir(folder):
+    img_path = os.path.join(folder, filename)
+    examples.append([img_path])
+
+random.shuffle(examples)
+
+
+iface = gr.Interface(
+    fn=classify,
+    title="Fashion MNIST Classifier - TAECAC @ FEUP",
+    description="Use an image to classify using the different Fashion MNIST model.",
+    inputs=gr.Image(label="Image", type="filepath"),
+    outputs=gr.Textbox(label="Classification output"),
+    examples=examples,
+    examples_per_page=100,  
+    theme=gr.themes.Soft(
+        primary_hue=gr.themes.colors.indigo,
+        secondary_hue=gr.themes.colors.gray,
+        neutral_hue=gr.themes.colors.slate,
+        font=["avenir"],
+    ),
+)
+
+iface.launch()

fmt.sh

@@ -0,0 +1,5 @@
+echo "Running fmt"
+echo "Running isort"
+isort . --profile black
+echo "Running black"
+black .