added prediction functionality to app and updated dockerfile

.dockerignore

@@ -1,3 +1,4 @@
 *.git
 **/__pycache__
+*.pt
 venv

Dockerfile

@@ -12,6 +12,8 @@ RUN pip install --upgrade pip \
 
 COPY . .
 
+RUN python download_model.py
+
 EXPOSE 3000
 
-CMD ["shiny", "run", "shinyapp/app.py", "-h", "0.0.0.0", "-p", "3000"]
+CMD ["shiny", "run", "app.py", "-h", "0.0.0.0", "-p", "3000"]

app.py

@@ -0,0 +1,78 @@
+from shiny import App, reactive, render, ui
+from shiny.types import ImgData
+
+from PIL import Image
+
+import torch
+import torchvision.transforms as transforms
+import modules.model as model
+
+def forward_prop(image_path):
+    imsize = 128
+    classes = ('Female', 'Male')
+    modelsave_name = 'model_parameters.pt'
+
+    if torch.backends.mps.is_available():
+        device = torch.device('mps')
+        device_name = 'Apple Silicon GPU'
+    elif torch.cuda.is_available():
+        device = torch.device('cuda')
+        device_name = 'CUDA'
+    else:
+        device = torch.device('cpu')
+        device_name = 'CPU'
+
+    torch.set_default_device(device)
+
+    resnet = model.resnetModel_128()
+    resnet.load_state_dict(torch.load(modelsave_name, map_location=device))
+    resnet.eval()
+
+    loader = transforms.Compose([
+        transforms.Resize([imsize, imsize]),
+        transforms.Grayscale(1),
+        transforms.ToTensor(),
+        transforms.Normalize(0, 1)
+    ])
+
+    image = Image.open(image_path).convert('RGB')
+    image_tensor = loader(image)
+    image_tensor = image_tensor.unsqueeze(0)
+
+    X = image_tensor.to(device)
+    y_pred = resnet.forward(X)
+    predicted = torch.max(y_pred.data,1)[1]
+
+    return f'Prediction: {classes[predicted]} with weight {y_pred[0][predicted].item()}. Predicted using {device_name}.'
+
+
+app_ui = ui.page_fluid(
+    ui.panel_title('Image Uploader'),
+    ui.input_file('image', 'Image', accept=['.png', '.jpg', '.jpeg']),
+    ui.output_image('show_image'),
+    ui.input_action_button('predict_gender', 'Predict'),
+    ui.output_text('prediction')
+)
+
+def server(input, output, session):
+    @render.image
+    def show_image():
+        if input.image() is None:
+            return None
+            
+        image_path = input.image()[0]['datapath']
+        img: ImgData = {'src': image_path}
+        return img
+    
+    @render.text
+    @reactive.event(input.predict_gender)
+    def prediction():
+        if input.image() is None:
+            return None
+        
+        image_path = input.image()[0]['datapath']
+        prediction = forward_prop(image_path)
+
+        return prediction
+
+app = App(app_ui, server)

download_model.py

@@ -0,0 +1,7 @@
+import os
+import gdown
+
+modelsave_name = 'model_parameters.pt'
+if os.path.isfile(modelsave_name) == False:
+    url = 'https://drive.google.com/file/d/1_mYn2LrhG080Xvt26tWBtJ8U_0F2E1-s/view?usp=sharing'
+    gdown.download(url, output=modelsave_name, fuzzy=True)

shinyapp/app.py

@@ -1,20 +0,0 @@
-from shiny import App, render, ui
-from shiny.types import ImgData
-
-app_ui = ui.page_fluid(
-    ui.panel_title('Image Uploader'),
-    ui.input_file('image', 'Image', multiple=True, accept=['.png', '.jpg', '.jpeg']),
-    ui.output_image('show_image')
-)
-
-def server(input, output, session):
-    @render.image
-    def show_image():
-        if input.image() is None:
-            return None
-            
-        image_path = input.image()[0]['datapath']
-        img: ImgData = {'src': image_path}
-        return img
-
-app = App(app_ui, server)