Initial Build.

app.py

@@ -0,0 +1,53 @@
+import torch
+import gradio as gr
+
+from model import create_effnetb2_model
+from pathlib import Path
+from timeit import default_timer as timer
+from typing import Tuple, Dict
+
+with open('class_names.txt', 'r') as f:
+    CLASS_NAMES = [l.strip() for l in f.readlines()]
+NUM_CLASSES = len(CLASS_NAMES)
+DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
+MODEL, MODEL_TRANSFORMS = create_effnetb2_model(NUM_CLASSES, load_st_dict=True)
+
+
+def predict(img) -> Tuple[Dict, float]:
+    start = timer()
+    X = MODEL_TRANSFORMS(img).unsqueeze(dim=0).to(DEVICE)
+    MODEL.eval()
+    with torch.inference_mode():
+        y_logits = MODEL(X)
+        y_prob = torch.softmax(y_logits, dim=1)
+    # Float casting due to Gradio's assumption that numpy objects
+    # should be iterated. Running `tolist()` prior to this
+    # dictionary comprehension can also fix this behavior.
+    y_prob = y_prob.squeeze(dim=0).cpu().numpy()
+    pred = {c: float(prob) for c, prob in zip(CLASS_NAMES, y_prob)}
+    end = timer()
+    return pred, end - start
+
+
+title = "FoodVision Big 🍖❓"
+description = (
+    "An [EfficientNetB2](https://pytorch.org/vision/stable/models/generated/torchvision.models.efficientnet_b2.html) "
+    "feature extraction that can classify images of 101 classes of food images."
+)
+article = (
+    "Created by [DiabeticOwl](https://huggingface.co/DiabeticOwl). "
+    "Uses the [Food101 dataset](https://data.vision.ee.ethz.ch/cvl/datasets_extra/food-101/)."
+)
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Image(shape=(288, 288), type='pil'),
+    outputs=[gr.Label(num_top_classes=5, label='Predictions'),
+             gr.Number(label='Prediction run time (s)')],
+    examples=list(Path('examples').glob('*.jpg')),
+    title=title,
+    description=description,
+    article=article
+)
+
+if __name__ == '__main__':
+    demo.launch()

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

model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6929a68394c7aef94ef492460a179fb54f62ef3645ebe8f27e6ad464d6080c84
+size 31850487

model.py

@@ -0,0 +1,32 @@
+import torch
+
+from pathlib import Path
+from torch import nn
+from torchvision.models import efficientnet_b2, EfficientNet_B2_Weights
+from typing import Optional, Tuple
+
+DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
+
+
+def create_effnetb2_model(
+    num_classes: int,
+    seed: Optional[int] = 42,
+    load_st_dict: Optional[bool] = False
+) -> Tuple[nn.Module, nn.Module]:
+    torch.manual_seed(seed)
+    torch.cuda.manual_seed(seed)
+    weights = EfficientNet_B2_Weights.DEFAULT
+    transforms = weights.transforms()
+    model = efficientnet_b2(weights=weights)
+    
+    model.classifier = nn.Sequential(
+      nn.Dropout(p=0.3, inplace=True),
+      nn.Linear(in_features=1408, out_features=num_classes, bias=True)
+    ).to(DEVICE)
+    if load_st_dict:
+        st_dict = Path('model.pth')
+        model.load_state_dict(torch.load(st_dict, map_location=DEVICE))
+    for param in model.parameters():
+        param.requires_grad = False
+    
+    return model.to(DEVICE), transforms

requirements.txt

@@ -0,0 +1,3 @@
+torch>=2.0.0
+torchvision>=0.15.0
+gradio>=3.33.1