app.py

# ── Gradio 설치 ──────────────────────────

import gradio as gr
import torch
import torch.nn as nn
import torchvision.transforms as transforms
from PIL import Image
from datasets import load_dataset

# ── 클래스 이름 가져오기 ──────────────────
dataset = load_dataset("food101", split="train[:1%]")
class_names = dataset.features['label'].names

# ── 모델 정의 (똑같이 붙여넣기) ───────────
class BottleneckBlock(nn.Module):
    expansion = 4

    def __init__(self, in_channels, mid_channels, stride=1):
        super().__init__()
        self.conv1 = nn.Conv2d(in_channels, mid_channels, kernel_size=1, bias=False)
        self.bn1   = nn.BatchNorm2d(mid_channels)
        self.conv2 = nn.Conv2d(mid_channels, mid_channels, kernel_size=3,
                               stride=stride, padding=1, bias=False)
        self.bn2   = nn.BatchNorm2d(mid_channels)
        self.conv3 = nn.Conv2d(mid_channels, mid_channels * self.expansion,
                               kernel_size=1, bias=False)
        self.bn3   = nn.BatchNorm2d(mid_channels * self.expansion)
        self.relu  = nn.ReLU(inplace=True)
        self.shortcut = nn.Sequential()
        if stride != 1 or in_channels != mid_channels * self.expansion:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_channels, mid_channels * self.expansion,
                          kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(mid_channels * self.expansion)
            )

    def forward(self, x):
        identity = x
        out = self.relu(self.bn1(self.conv1(x)))
        out = self.relu(self.bn2(self.conv2(out)))
        out = self.bn3(self.conv3(out))
        out += self.shortcut(identity)
        out = self.relu(out)
        return out

class ResNet50(nn.Module):
    def __init__(self, num_classes=101):
        super().__init__()
        self.conv1   = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False)
        self.bn1     = nn.BatchNorm2d(64)
        self.relu    = nn.ReLU(inplace=True)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1  = self._make_layer( 64,  64, blocks=3, stride=1)
        self.layer2  = self._make_layer(256, 128, blocks=4, stride=2)
        self.layer3  = self._make_layer(512, 256, blocks=6, stride=2)
        self.layer4  = self._make_layer(1024, 512, blocks=3, stride=2)
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc      = nn.Linear(2048, num_classes)

    def _make_layer(self, in_channels, mid_channels, blocks, stride):
        layers = [BottleneckBlock(in_channels, mid_channels, stride=stride)]
        for _ in range(1, blocks):
            layers.append(BottleneckBlock(mid_channels * 4, mid_channels))
        return nn.Sequential(*layers)

    def forward(self, x):
        x = self.maxpool(self.relu(self.bn1(self.conv1(x))))
        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)
        x = self.avgpool(x)
        x = torch.flatten(x, 1)
        x = self.fc(x)
        return x

# ── 모델 불러오기 ─────────────────────────
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

model = ResNet50(num_classes=101)
model.load_state_dict(
    torch.load('resnet50_food101.pth', map_location=device)
)
model = model.to(device)
model.eval()

# ── 전처리 ────────────────────────────────
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    )
])

# ── 예측 함수 ─────────────────────────────
def predict(image):
    # PIL 이미지 → 텐서 변환
    img_tensor = transform(image).unsqueeze(0).to(device)
    # unsqueeze(0) = (3,224,224) → (1,3,224,224) 배치 차원 추가

    with torch.no_grad():
        output = model(img_tensor)
        probs  = torch.softmax(output, dim=1)  # 점수 → 확률
        top5   = probs.topk(5)                 # 상위 5개

    # 결과 딕셔너리로 반환
    result = {}
    for i in range(5):
        label = class_names[top5.indices[0][i].item()]
        prob  = top5.values[0][i].item()
        result[label] = prob

    return result

# ── Gradio 인터페이스 ──────────────────────
demo = gr.Interface(
    fn=predict,                          # 예측 함수
    inputs=gr.Image(type="pil"),         # 입력: 이미지
    outputs=gr.Label(num_top_classes=5), # 출력: 상위 5개 클래스
    title="🍔 Food-101 분류기",
    description="음식 사진을 올리면 어떤 음식인지 맞춰줘요! (101가지 음식 분류)",
    examples=[],                         # 예시 이미지 (있으면 추가)
)

demo.launch()