app.py

from fastai.vision.all import *
from icevision.all import *
import gradio as gr
import torch

# Definimos una función que se encarga de llevar a cabo las predicciones
def predict(img):
    class_map = ClassMap(['Kangaroo'])
    model = models.torchvision.faster_rcnn.model(backbone=models.torchvision.faster_rcnn.backbones.resnet18_fpn(pretrained=True), num_classes=len(class_map))
    # model = models.torchvision.faster_rcnn.model(backbone=models.torchvision.faster_rcnn.backbones.resnet18_fpn, num_classes=len(class_map))
    state_dict = torch.load('model.pth', map_location=torch.device('cpu'))
    model.load_state_dict(state_dict)
    
    infer_tfms = tfms.A.Adapter([*tfms.A.resize_and_pad(384),tfms.A.Normalize()])
    
    img = PILImage.create(img)
    pred_dict  = models.torchvision.faster_rcnn.end2end_detect(img, infer_tfms, model.to("cpu"), class_map=class_map, detection_threshold=0.5)
    return pred_dict['img']
    
# Creamos la interfaz y la lanzamos. 
gr.Interface(fn=predict, inputs=gr.inputs.Image(shape=(128, 128)), outputs=gr.outputs.Image(),examples=['kangarooc.jpg']).launch(share=False)