app.py

import os
import torch
from torch.utils.data import Dataset, DataLoader
import pandas as pd
import numpy as np
import joblib
from io import BytesIO
from PIL import Image
from torchvision import transforms,models
from sklearn.preprocessing import LabelEncoder,MinMaxScaler
from gradio import Interface, Image, Label, HTML
from huggingface_hub import snapshot_download

# Retrieve the token from the environment variables
token = os.environ.get("token")

# Download the repository snapshot
local_dir = snapshot_download(
    repo_id="robocan/GeoG_coordinate",
    repo_type="model",
    local_dir="SVD",
    token=token
)

device = 'cpu'
le = LabelEncoder()
le = joblib.load("SVD/le.gz")
MMS = joblib.load("SVD/MMS.gz")
len_classes = len(le.classes_) + 1

class ModelPre(torch.nn.Module):
    def __init__(self):
        super().__init__()
        self.embedding = torch.nn.Sequential(
            *list(models.convnext_small(weights=models.ConvNeXt_Small_Weights.IMAGENET1K_V1).children())[:-1],
            torch.nn.Flatten(),
            torch.nn.Linear(in_features=768,out_features=512),
            torch.nn.ReLU(),
            torch.nn.Linear(in_features=512,out_features=len_classes),
        )
        # Freeze all layers

    def forward(self, data):
        return self.embedding(data)
    
# Load the pretrained model 
model = ModelPre()
#for param in model.parameters():
#   param.requires_grad = False
class GeoGcord(torch.nn.Module):
    def __init__(self):
        super().__init__()
        self.embedding = torch.nn.Sequential(
            *list(model.children())[0][:-1],
            torch.nn.Linear(in_features=512,out_features=256),
            torch.nn.ReLU(),
            torch.nn.Linear(in_features=256,out_features=128),
            torch.nn.ReLU(),
            torch.nn.Linear(in_features=128,out_features=2),
        )
        # Freeze all layers

    def forward(self, data):
        return self.embedding(data)
    
    

# Load the pre-trained model 
model = GeoGcord()
model_w = torch.load("SVD/GeoG.pth", map_location=torch.device(device))

model.load_state_dict(model_w['model'])

cmp = transforms.Compose([
    transforms.ToTensor(),
    transforms.Resize(size=(224, 224), antialias=True),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])

# Predict function for the new regression model
def predict(input_img):
    with torch.inference_mode():
        img = cmp(input_img).unsqueeze(0)
        res = model(img.to(device))
        # Assuming res is a 2-layer regression output, and MMS.inverse_transform is needed
        prediction = MMS.inverse_transform(res.cpu().numpy()).flatten()
        return prediction

# Function to generate Plotly map figure
def create_map_figure(lat, lon):
    fig = go.Figure(go.Scattermapbox(
        lat=[lat],
        lon=[lon],
        mode='markers',
        marker=go.scattermapbox.Marker(
            size=14
        ),
        text=[f'Lat: {lat}, Lon: {lon}'],
        hoverinfo='text'
    ))

    fig.update_layout(
        mapbox_style="open-street-map",
        hovermode='closest',
        mapbox=dict(
            bearing=0,
            center=go.layout.mapbox.Center(
                lat=lat,
                lon=lon
            ),
            pitch=0,
            zoom=10
        ),
    )

    return fig

# Create label output function
def create_label_output(predictions):
    lat, lon = predictions
    fig = create_map_figure(lat, lon)
    return fig

# Predict and plot function
def predict_and_plot(input_img):
    predictions = predict(input_img)
    return create_label_output(predictions)

# Gradio app definition
with gr.Blocks() as gradio_app:
    with gr.Column():
        input_image = gr.Image(label="Upload an Image", type="pil")
        output_map = gr.Plot(label="Predicted Location on Map")
        btn_predict = gr.Button("Predict")

        btn_predict.click(predict_and_plot, inputs=input_image, outputs=output_map)

    gradio_app.launch()