Spaces:
Build error
Build error
File size: 994 Bytes
f7dfb0f 9c52842 259984c f7dfb0f 9c52842 cc1b998 f7dfb0f d995000 259984c 9c52842 ae16aa7 9c52842 ae16aa7 03cf1d4 fc466bc 03cf1d4 abbfcd2 9c52842 03cf1d4 ae16aa7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 |
import gradio as gr
import tensorflow as tf
import numpy as np
num_classes = 200
IMG_HEIGHT = 300
IMG_WIDTH = 300
with open("classlabel.txt", 'r') as file:
CLASS_LABEL = [x.strip() for x in file.readlines()]
def normalize_image(img):
img = tf.cast(img, tf.float32) / 255.
img = tf.image.resize(img, (IMG_HEIGHT, IMG_WIDTH), method='bilinear')
return img
def predict_top_classes(img, num_top_classes=5):
img = img.convert('RGB')
img_data = normalize_image(img)
x = np.array(img_data)
x = np.expand_dims(x, axis=0)
temp = model.predict(x)
idx = np.argsort(np.squeeze(temp))[::-1]
top5_value = np.asarray([temp[0][i] for i in idx[0:5]])
top5_idx = idx[0:5]
return {CLASS_LABEL[i]:str(v) for i,v in zip(top5_idx,top5_value)}
model = tf.keras.models.load_model("Xception.h5")
interface = gr.Interface(predict_top_classes, gr.inputs.Image(type='pil'), outputs='label', args={'num_top_classes': 5})
interface.launch()
|