app.py

import gradio as gr

import torch
import torch.nn.functional as F
from torchvision import transforms

# load the model
model = torch.jit.load("food_classifier_resnet18.ptl")

# Transformations that will be applied
the_transform = transforms.Compose([
        transforms.Resize((224,224)),
        transforms.CenterCrop((224,224)),
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.485,0.456,0.406],std=[0.229,0.224,0.225])
])

# Classes 
class_names = ['Apple Pie','Bibimbap','Cannoli','Edamame','Falafel','French Toast','Ice Cream','Ramen','Sushi','Tiramisu']

# Returns transformed image
def transform_img(img):
    return the_transform(img) 

# Returns string with class and probability
def classify_img(img):
    # Applying transformation to the image
    model_img = transform_img(img)
    model_img = model_img.view(1,3,224,224)
    
    # Running image through the model
    model.eval()
    with torch.no_grad():
        result = model(model_img)
    
    # Converting values to softmax values 
    result = F.softmax(result,dim=1)
    # Grabbing top 3 indices and probabilities for each index
    top3_prob, top3_catid = torch.topk(result,3)
    
    # Dictionary I will display
    model_output = {}
    for i in range(top3_prob.size(1)):
        
        model_output[class_names[top3_catid[0][i].item()]] = top3_prob[0][i].item()
    print(model_output)
    return model_output 


demo = gr.Interface(classify_img,
                        inputs = gr.inputs.Image(type="pil"), 
                        outputs = gr.outputs.Label(type="confidences",num_top_classes=3),
                        title = "Food Classifier!", 
                        description = "Insert food image you would like to classify! Returns confidence % for the top three categories <br> Categories: Apple Pie, Bibimbap, Cannoli, Edamame, Falafel, French Toast, Ice Cream, Ramen, Sushi, Tiramisu",
                    )

demo.launch(inline=False)