|
import os |
|
import json |
|
|
|
import torch |
|
from PIL import Image |
|
from torchvision import transforms |
|
import matplotlib.pyplot as plt |
|
|
|
from model import efficientnetv2_m as create_model |
|
|
|
|
|
|
|
def predict(img): |
|
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") |
|
|
|
img_size = {"s": [300, 384], |
|
"m": [384, 480], |
|
"l": [384, 480]} |
|
num_model = "m" |
|
|
|
data_transform = transforms.Compose( |
|
[transforms.Resize(img_size[num_model][1]), |
|
transforms.CenterCrop(img_size[num_model][1]), |
|
transforms.ToTensor(), |
|
transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])]) |
|
img = data_transform(img) |
|
|
|
img = torch.unsqueeze(img, dim=0) |
|
json_path = './class_indices.json' |
|
json_file = open(json_path, "r") |
|
class_indict = json.load(json_file) |
|
model = create_model(num_classes=5).to(device) |
|
model_weight_path = "./weights/model-20.pth" |
|
model.load_state_dict(torch.load(model_weight_path, map_location=device)) |
|
model.eval() |
|
with torch.no_grad(): |
|
|
|
output = torch.squeeze(model(img.to(device))).cpu() |
|
predict = torch.softmax(output, dim=0) |
|
predict_cla = torch.argmax(predict).numpy() |
|
print_res = "class: {} \n prob: {:.3}".format(class_indict[str(predict_cla)], |
|
predict[predict_cla].numpy()) |
|
return print_res |
|
|
|
import gradio as gr |
|
examples = ['d.jpg', 'rose.jpg', 'rose2.jpg', 'images.jpg'] |
|
inter = gr.Interface(fn=predict, |
|
inputs=gr.inputs.Image(type="pil"), |
|
outputs=gr.outputs.Label(num_top_classes=5), |
|
title = 'Five types of flower Detection', |
|
description= 'This program can be used to detect five types of flowers: "daisy", "dandelion", "roses", "sunflowers", "tulips", and the program will give the classification results along with a confidence score.', theme = 'huggingface') |
|
inter.launch(inline=False,debug=True) |
|
|
