import gradio as gr
import torch
import torch.nn as nn
from torchvision.models import resnet50, ResNet50_Weights
from torchvision import transforms
from PIL import Image


class Vocabulary:
    def __init__(self):
        self.itos = {}
        self.stoi = {}

    def load(self, stoi, itos):
        self.stoi = stoi
        self.itos = itos


class EncoderCNN(nn.Module):
    def __init__(self, embed_size):
        super(EncoderCNN, self).__init__()
        resnet = resnet50(weights=ResNet50_Weights.DEFAULT)
        modules = list(resnet.children())[:-1]
        self.resnet = nn.Sequential(*modules)
        self.linear = nn.Linear(resnet.fc.in_features, embed_size)
        self.bn = nn.BatchNorm1d(embed_size, momentum=0.01)

    def forward(self, images):
        with torch.no_grad():
            features = self.resnet(images) 
        features = features.view(features.size(0), -1)  
        features = self.linear(features)                
        features = self.bn(features)                   
        return features


class DecoderRNN(nn.Module):
    def __init__(self, embed_size, hidden_size, vocab_size, num_layers=1):
        super(DecoderRNN, self).__init__()
        self.embed = nn.Embedding(vocab_size, embed_size)
        self.lstm = nn.LSTM(embed_size, hidden_size, num_layers, batch_first=True)
        self.linear = nn.Linear(hidden_size, vocab_size)

    def forward(self, features, captions):
        embeddings = self.embed(captions)
        inputs = torch.cat((features.unsqueeze(1), embeddings), 1)
        hiddens, _ = self.lstm(inputs)
        outputs = self.linear(hiddens)
        return outputs

    def sample(self, features, vocab, max_len=30):
        output_ids = []
        states = None

        inputs = features.unsqueeze(1)

        for _ in range(max_len):
            hiddens, states = self.lstm(inputs, states)
            outputs = self.linear(hiddens.squeeze(1))
            predicted = outputs.argmax(1)
            output_ids.append(predicted.item())

            if vocab.itos[predicted.item()] == "<end>":
                break

            inputs = self.embed(predicted).unsqueeze(1)

        return output_ids

checkpoint_path = "./checkpoints/caption_model_epoch30.pth"

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

embed_size = 256
hidden_size = 512
num_layers = 1

checkpoint = torch.load(checkpoint_path, map_location=device)

vocab = Vocabulary()
vocab.load(checkpoint['vocab_stoi'], checkpoint['vocab_itos'])
vocab_size = len(vocab.stoi)

encoder = EncoderCNN(embed_size).to(device)
decoder = DecoderRNN(embed_size, hidden_size, vocab_size, num_layers).to(device)

encoder.load_state_dict(checkpoint['encoder_state_dict'])
decoder.load_state_dict(checkpoint['decoder_state_dict'])

encoder.eval()
decoder.eval()

transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor()
])

def generate_caption(image):
    image = Image.fromarray(image).convert("RGB")
    image = transform(image).unsqueeze(0).to(device)

    with torch.no_grad():
        features = encoder(image)          
        output_ids = decoder.sample(features, vocab)

    caption = []
    for idx in output_ids:
        word = vocab.itos[idx]
        if word == "<end>":
            break
        caption.append(word)

    return ' '.join(caption)

demo = gr.Interface(
    fn=generate_caption,
    inputs=gr.Image(type="numpy"),
    outputs="text",
    title="Skin Disease Image Captioning",
    description="Upload an image of a skin disease to generate a descriptive caption using your trained model."
)

if __name__ == "__main__":
    demo.launch()