initial commit

app.py

@@ -0,0 +1,73 @@
+
+"""The main parts are:
+1. Imports and class names setup
+2. Model and transforms preparation
+3. Write a predict function for gradio to use
+4. Write the Gradio app and the launch command
+"""
+import os
+from typing import Tuple, Dict, List
+import PIL
+import torch
+from torch import nn
+import torchvision
+import gradio as gr
+from timeit import default_timer as timer
+from model import create_effnetb2_model
+
+# Setup class names
+filepath = "./class_names.txt"
+with open(filepath, "r") as f:
+    class_names = [food_name.strip() for food_name in f.readlines()]
+    
+model, transforms = create_effnetb2_model(num_classes = len(class_names))
+
+# Load saved weights into the model, and load the model onto the CPU
+model.load_state_dict(torch.load(f = "effnetb2_feature_extractor_food101_20_percent.pth", 
+                                 map_location = torch.device('cpu')))
+
+# Write function to run inference on gradio
+def predict(img: PIL.Image, 
+            model: nn.Module = model,
+            transforms: torchvision.transforms = transforms,
+            class_names: List[str] = class_names) -> Tuple[Dict, float]:
+    """Function to predict image class on gradio
+
+    Args:
+        img (np.array): Image as a numpy array
+        model (nn.Module, optional): Model. Defaults to vit.
+        class_names (List[str], optional): List of class anmes. Defaults to class_names.
+
+    Returns:
+        Tuple[Dict, float]: Tuplefor further processing on gradio
+    """
+    start_time = timer()
+    img = transforms(img).unsqueeze(0) #add batch dimension
+    model.eval()
+    with torch.inference_mode():
+        pred_probs = torch.softmax(model(img), dim = 1)
+    pred_labels_and_probs = {class_names[i]: float(pred_probs[0][i]) for i in range(len(class_names))}
+    end_time = timer()
+    pred_time = round(end_time - start_time, 4)
+    return pred_labels_and_probs, pred_time
+
+# Create example_list
+example_list = [["examples/" + example] for example in os.listdir("examples")]
+
+# Create Gradio App
+title = 'FoodVision 101 🍔🍣🥩🍕🍦😋'
+description = "Using an [EfficientNet](https://arxiv.org/abs/1905.11946) for Image Classification of 101 different food classes"
+article = "Created by [Titus Lim](https://github.com/tituslhy)"
+
+demo = gr.Interface(fn = predict,
+                    inputs = gr.Image(type = "pil"),
+                    outputs = [gr.Label(num_top_classes = 5, label = "Predictions"),
+                               gr.Number(label = "Prediction time (s)")],
+                    examples 😋= example_list,
+                    title = title,
+                    description = description,
+                    article = article)
+
+# Launch demo
+demo.launch(debug = False, #prints errors locally
+            )

class_names.txt

@@ -0,0 +1,101 @@
+apple_pie
+baby_back_ribs
+baklava
+beef_carpaccio
+beef_tartare
+beet_salad
+beignets
+bibimbap
+bread_pudding
+breakfast_burrito
+bruschetta
+caesar_salad
+cannoli
+caprese_salad
+carrot_cake
+ceviche
+cheese_plate
+cheesecake
+chicken_curry
+chicken_quesadilla
+chicken_wings
+chocolate_cake
+chocolate_mousse
+churros
+clam_chowder
+club_sandwich
+crab_cakes
+creme_brulee
+croque_madame
+cup_cakes
+deviled_eggs
+donuts
+dumplings
+edamame
+eggs_benedict
+escargots
+falafel
+filet_mignon
+fish_and_chips
+foie_gras
+french_fries
+french_onion_soup
+french_toast
+fried_calamari
+fried_rice
+frozen_yogurt
+garlic_bread
+gnocchi
+greek_salad
+grilled_cheese_sandwich
+grilled_salmon
+guacamole
+gyoza
+hamburger
+hot_and_sour_soup
+hot_dog
+huevos_rancheros
+hummus
+ice_cream
+lasagna
+lobster_bisque
+lobster_roll_sandwich
+macaroni_and_cheese
+macarons
+miso_soup
+mussels
+nachos
+omelette
+onion_rings
+oysters
+pad_thai
+paella
+pancakes
+panna_cotta
+peking_duck
+pho
+pizza
+pork_chop
+poutine
+prime_rib
+pulled_pork_sandwich
+ramen
+ravioli
+red_velvet_cake
+risotto
+samosa
+sashimi
+scallops
+seaweed_salad
+shrimp_and_grits
+spaghetti_bolognese
+spaghetti_carbonara
+spring_rolls
+steak
+strawberry_shortcake
+sushi
+tacos
+takoyaki
+tiramisu
+tuna_tartare
+waffles

effnetb2_feature_extractor_food101_20_percent.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8edf12deee22fbca3e7cd777df1130c572017a1cc6677ddb9c6a3d3053df29b5
+size 31841405

model.py

@@ -0,0 +1,18 @@
+
+import torch
+import torchvision
+from torch import nn
+
+def create_effnetb2_model(num_classes: int = 101,
+                          seed: int = 42):
+    effnetb2_weights = torchvision.models.EfficientNet_B2_Weights.DEFAULT
+    effnetb2_transforms = effnetb2_weights.transforms()
+    effnetb2 = torchvision.models.efficientnet_b2(weights=effnetb2_weights)
+    for param in effnetb2.parameters():
+        param.requires_grad = False
+    torch.manual_seed(seed)
+    effnetb2.classifier = nn.Sequential(
+        nn.Dropout(p = 0.3, inplace = True),
+        nn.Linear(in_features = 1408, out_features = num_classes)
+    )
+    return effnetb2, effnetb2_transforms

requirements.txt

@@ -0,0 +1,4 @@
+gradio==3.41.2
+Pillow==9.5.0
+torch==2.0.1
+torchvision==0.15.2