app.py

from huggingface_hub import from_pretrained_fastai
import gradio as gr
from fastai.vision.all import *

def get_y_fn (x):
    return Path(str(x).replace("Images","Labels").replace("color","gt").replace(".jpg",".png"))

from albumentations import (
    Compose,
    OneOf,
    ElasticTransform,
    GridDistortion,
    OpticalDistortion,
    HorizontalFlip,
    Rotate,
    Transpose,
    CLAHE,
    ShiftScaleRotate,
    RandomBrightnessContrast,
    GaussNoise
)

class SegmentationAlbumentationsTransform(ItemTransform):
    split_idx = 0

    def __init__(self, aug):
        self.aug = aug

    def encodes(self, x):
        img,mask = x
        aug = self.aug(image=np.array(img), mask=np.array(mask))
        return PILImage.create(aug["image"]), PILMask.create(aug["mask"])

class TargetMaskConvertTransform(ItemTransform):
    def __init__(self):
        pass
    def encodes(self, x):
        img,mask = x

        #Convert to array
        mask = np.array(mask)

        # Uvas
        mask[mask==255]=1

        # Hojas
        mask[mask==150]=2

        # Poste
        mask[mask==76]=3
        mask[mask==74]=3

        # Madera
        mask[mask==29]=4
        mask[mask==25]=4

        # Back to PILMask
        mask = PILMask.create(mask)
        return img, mask

repo_id = "ancebuc/grapes-segmentation"

learner = from_pretrained_fastai(repo_id)

aux = learner.model
aux = aux.cpu()

import torchvision.transforms as transforms
img = PILImage.create('color_158.jpg')
transformer=transforms.Compose([transforms.Resize((480,640)),
                                transforms.ToTensor(),
                                transforms.Normalize(
                                                    [0.485, 0.456, 0.406],
                                                    [0.229, 0.224, 0.225])])
img=transformer(img).unsqueeze(0)
img=img.cpu()

model=torch.jit.trace(aux, (img))

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

model.to(device)

import torchvision.transforms as transforms
def transform_image(image):
    my_transforms = transforms.Compose([transforms.ToTensor(),
                                        transforms.Normalize(
                                            [0.485, 0.456, 0.406],
                                            [0.229, 0.224, 0.225])])
    image_aux = image
    return my_transforms(image_aux).unsqueeze(0).to(device)

# Definimos una función que se encarga de llevar a cabo las predicciones
def predict(img):
    img = PILImage.create(img)

    image = transforms.Resize((480,640))(img)
    tensor = transform_image(image=image)

    with torch.no_grad():
        outputs = model(tensor)

    outputs = torch.argmax(outputs,1)

    mask = np.array(outputs.cpu())
    mask = np.reshape(mask,(480,640))

    # Añadimos una dimesionalidad para colocar color
    mask = np.expand_dims(mask, axis=2)

    # Y añadimos los tres canales
    mask = np.repeat(mask, 3, axis=2)

    # Creamos las máscaras
    uvas = np.all(mask == [1, 1, 1], axis=2)
    hojas = np.all(mask == [2, 2, 2], axis=2)
    poste = np.all(mask == [3, 3, 3], axis=2)
    madera = np.all(mask == [4, 4, 4], axis=2)
    
    # Uvas
    mask[uvas] = [255, 255, 255]
    
    # Hojas
    mask[hojas] = [0, 255, 0]
    
    # Poste
    mask[poste] = [0, 0, 255]
    
    # Madera
    mask[madera] = [255, 0, 0]

    return Image.fromarray(mask.astype('uint8'))
    
# Creamos la interfaz y la lanzamos. 
gr.Interface(fn=predict, inputs="image", outputs="image",examples=['color_158.jpg','color_157.jpg']).launch(share=False)