app.py

import streamlit as st
import torch
import numpy as np
from PIL import Image
import rasterio
from rasterio.windows import Window
from tqdm.auto import tqdm
import io
import zipfile
import os
import albumentations as albu
import segmentation_models_pytorch as smp
from albumentations.pytorch.transforms import ToTensorV2
import geopandas as gpd
from shapely.geometry import shape
from shapely.ops import unary_union
from rasterio.features import shapes


DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
ENCODER = 'se_resnext50_32x4d'
ENCODER_WEIGHTS = 'imagenet'


# Load and prepare the model
@st.cache_resource
def load_model():
    model = torch.load('deeplabv3 v15.pth', map_location=DEVICE)
    model.eval().float()
    return model


best_model = load_model()


def to_tensor(x, **kwargs):
    return x.astype('float32')


# Preprocessing
preprocessing_fn = smp.encoders.get_preprocessing_fn(ENCODER, ENCODER_WEIGHTS)


def get_preprocessing():
    _transform = [
        albu.Resize(512, 512),
        albu.Lambda(image=preprocessing_fn),
        albu.Lambda(image=to_tensor, mask=to_tensor),
        ToTensorV2(),
    ]
    return albu.Compose(_transform)


preprocess = get_preprocessing()


@torch.no_grad()
def process_and_predict(image, model):
    if isinstance(image, Image.Image):
        image = np.array(image)

    if image.ndim == 2:
        image = np.stack([image] * 3, axis=-1)
    elif image.shape[2] == 4:
        image = image[:, :, :3]

    preprocessed = preprocess(image=image)['image']
    input_tensor = preprocessed.unsqueeze(0).to(DEVICE)

    mask = model(input_tensor)
    mask = torch.sigmoid(mask)
    mask = (mask > 0.6).float()

    mask_image = Image.fromarray((mask.squeeze().cpu().numpy() * 255).astype(np.uint8))

    return mask_image


def extract_tiles(map_file, model, tile_size=512, overlap=0, batch_size=4,threshold=0.6):
    tiles = []

    with rasterio.open(map_file) as src:
        height = src.height
        width = src.width

        effective_tile_size = tile_size - overlap

        for y in tqdm(range(0, height, effective_tile_size)):
            for x in range(0, width, effective_tile_size):
                batch_images = []
                batch_metas = []

                for i in range(batch_size):
                    curr_y = y + (i * effective_tile_size)
                    if curr_y >= height:
                        break

                    window = Window(x, curr_y, tile_size, tile_size)
                    out_image = src.read(window=window)

                    if out_image.shape[0] == 1:
                        out_image = np.repeat(out_image, 3, axis=0)
                    elif out_image.shape[0] != 3:
                        raise ValueError("The number of channels in the image is not supported")

                    out_image = np.transpose(out_image, (1, 2, 0))
                    tile_image = Image.fromarray(out_image.astype(np.uint8))

                    out_meta = src.meta.copy()
                    out_meta.update({
                        "driver": "GTiff",
                        "height": tile_size,
                        "width": tile_size,
                        "transform": rasterio.windows.transform(window, src.transform)
                    })
                    tile_image = np.array(tile_image)

                    preprocessed_tile = preprocess(image=tile_image)['image']
                    batch_images.append(preprocessed_tile)
                    batch_metas.append(out_meta)

                if not batch_images:
                    break

                batch_tensor = torch.cat([img.unsqueeze(0).to(DEVICE) for img in batch_images], dim=0)
                with torch.no_grad():
                    batch_masks = model(batch_tensor.to(DEVICE))

                batch_masks = torch.sigmoid(batch_masks)
                batch_masks = (batch_masks > threshold).float()

                for j, mask_tensor in enumerate(batch_masks):
                    mask_resized = torch.nn.functional.interpolate(mask_tensor.unsqueeze(0),
                                                                   size=(tile_size, tile_size), mode='bilinear',
                                                                   align_corners=False).squeeze(0)

                    mask_array = mask_resized.squeeze().cpu().numpy()

                    if mask_array.any() == 1:
                        tiles.append([mask_array, batch_metas[j]])

    return tiles






def create_vector_mask(tiles, output_path):
    all_polygons = []
    for mask_array, meta in tiles:
        # Ensure mask is binary
        mask_array = (mask_array > 0).astype(np.uint8)

        # Get shapes from the mask
        mask_shapes = list(shapes(mask_array, mask=mask_array, transform=meta['transform']))

        # Convert shapes to Shapely polygons
        polygons = [shape(geom) for geom, value in mask_shapes if value == 1]

        all_polygons.extend(polygons)
    # Perform union of all polygons
    union_polygon = unary_union(all_polygons)
    # Create a GeoDataFrame
    gdf = gpd.GeoDataFrame({'geometry': [union_polygon]}, crs=meta['crs'])
    # Save to file
    gdf.to_file(output_path)

    # Calculate area in square meters
    area_m2 = gdf.to_crs(epsg=3857).area.sum()
    # Convert to hectares
    area_ha = area_m2 / 10000

    return gdf, area_ha


def main():
    st.title("TIF File Processor")

    uploaded_file = st.file_uploader("Choose a TIF file", type="tif")

    if uploaded_file is not None:
        st.write("File uploaded successfully!")

        threshold= st.slider(
        'Select a float value',
        min_value=0.1,
        max_value=0.9,
        value=0.5,
        step=0.05
        )
        overlap= st.slider(
        'Select a float value',
        min_value=50,
        max_value=150,
        value=100,
        step=25
        )
        st.write('Selected threshold value:', threshold)
        st.write('Selected overlap value:', overlap)

        if st.button("Process File"):
            st.write("Processing...")

            with open("temp.tif", "wb") as f:
                f.write(uploaded_file.getbuffer())

            best_model.float()
            tiles = extract_tiles("temp.tif", best_model, tile_size=512, overlap=overlap, batch_size=4,threshold=threshold)

            st.write("Processing complete!")

            output_path = "output_mask.shp"
            result_gdf, area_ha = create_vector_mask(tiles, output_path)

            st.write("Vector mask created successfully!")
            st.write(f"Total area occupied by polygons: {area_ha:.2f} hectares")

            # Offer the shapefile for download
            shp_files = [f for f in os.listdir() if
                         f.startswith("output_mask") and f.endswith((".shp", ".shx", ".dbf", ".prj"))]

            with io.BytesIO() as zip_buffer:
                with zipfile.ZipFile(zip_buffer, 'a', zipfile.ZIP_DEFLATED, False) as zip_file:
                    for file in shp_files:
                        zip_file.write(file)

                zip_buffer.seek(0)
                st.download_button(
                    label="Download shapefile",
                    data=zip_buffer,
                    file_name="output_mask.zip",
                    mime="application/zip"
                )

            # Clean up temporary files
            os.remove("temp.tif")
            for file in shp_files:
                os.remove(file)


if __name__ == "__main__":
    main()