import gradio as gr
import PIL
import torch
from fastai.vision.all import models
from icevision.all import ClassMap, tfms


class_map = ClassMap(["racoon"])
state_dict = torch.load("fasterRCNNRaccoons.pth")
model = models.torchvision.faster_rcnn.model(
    backbone=models.torchvision.faster_rcnn.backbones.resnet50_fpn,
    num_classes=len(class_map),
)
model.load_state_dict(state_dict)

size = 384

def predict(img_file):
    img = PIL.Image.open(img_file)
    infer_tfms = tfms.A.Adapter([*tfms.A.resize_and_pad(size), tfms.A.Normalize()])
    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"]

gr.Interface(
    fn=predict,
    inputs=gr.inputs.Image(shape=(128, 128)),
    outputs=gr.outputs.Label(num_top_classes=3),
    examples=["raccoon-27.jpg", "raccoon-20.jpg"],
).launch(share=False)