Spaces:
Runtime error
Runtime error
File size: 4,767 Bytes
7b68e51 fac7e00 7b68e51 44df331 7b68e51 94da9e4 7b68e51 34c2d11 7b68e51 295f3ea 7b68e51 eb29ff3 7b68e51 eb29ff3 7b68e51 eb29ff3 7b68e51 3469024 7b68e51 44df331 7b68e51 9cd0c22 7b68e51 3469024 7b68e51 |
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 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 |
import tensorflow as tf
from matplotlib import pyplot as plt
from skimage.transform import rescale, resize
import pickle as pkl
import numpy as np
import os
import cv2
from PIL import Image,ImageFont, ImageDraw
import CALTextModel
import gradio as gr
#### training setup parameters ####
lambda_val=1e-4
gamma_val=1
os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
################################### Utility functions###################################
def load_dict_picklefile(dictFile):
fp=open(dictFile,'rb')
lexicon=pkl.load(fp)
fp.close()
return lexicon,lexicon[' ']
def preprocess_img(img):
if len(img.shape)>2:
img= cv2.cvtColor(img.astype('float32'), cv2.COLOR_BGR2GRAY)
height=img.shape[0]
width=img.shape[1]
if(width<300):
result = np.ones([img.shape[0], img.shape[1]*2])*255
result[0:img.shape[0],img.shape[1]:img.shape[1]*2]=img
img=result
img=cv2.resize(img, dsize=(800,100), interpolation = cv2.INTER_AREA)
img=(img-img.min())/(img.max()-img.min())
xx_pad = np.zeros((100, 800), dtype='float32')
xx_pad[:,:] =1
xx_pad = xx_pad[None, :, :]
img=img[None, :, :]
return img, xx_pad
worddicts,_ = load_dict_picklefile('vocabulary.pkl')
worddicts_r = [None] * len(worddicts)
i=1
for kk, vv in worddicts.items():
if(i<len(worddicts)):
worddicts_r[vv] = kk
else:
break
i=i+1
# Create an instance of the model
CALTEXT = CALTextModel.CALTEXT_Model(training=False)
CALTEXT.load_weights('final_caltextModel/cp-0037.ckpt')
test_loss = tf.keras.metrics.Mean(name='test_loss')
@tf.function(experimental_relax_shapes=True)
def execute_model(xx,xx_mask,CALTEXT):
anno = CALTEXT(xx,xx_mask, training=False)
hidden_state_0 = CALTEXT.get_hidden_state_0(anno)
return anno,hidden_state_0
def test_error( images, x_mask):
# training=False is only needed if there are layers with different
# behavior during training versus inference (e.g. Dropout).
batch_loss=0
img_ind=1
for img_ind in range(len(images)):
xx = images[img_ind][tf.newaxis, ... ]
xx_mask = x_mask[img_ind][tf.newaxis, ... ]
anno,hidden_state_0=execute_model(xx,xx_mask,CALTEXT)
sample, score,hypalpha=CALTextModel.get_sample(anno, hidden_state_0,10, 130, False, False, CALTEXT)
score = score / np.array([len(s) for s in sample])
ss = sample[score.argmin()]
img_ind=img_ind+1
ind=0
num=int(len(ss)/2)
#### output string
ind=0
outstr=u''
frames = []
font = ImageFont.truetype("Jameel Noori Nastaleeq.ttf",60)
while (ind<len(ss)-1):
k=(len(ss)-2)-ind
outstr=outstr+worddicts_r[int(ss[k])]
textimg = Image.new('RGB', (1400,100),(255,255,255))
drawtext = ImageDraw.Draw(textimg)
drawtext.text((20, 20), outstr ,(0,0,0),font=font)
fig,axes=plt.subplots(2,1)
axes[0].imshow(textimg)
axes[0].axis('off')
axes[1].axis('off')
axes[1].imshow(xx[0,:,:],cmap='gray')
visualization=resize(hypalpha[k], (100,800),anti_aliasing=True)
axes[1].imshow(255-(255 * visualization), alpha=0.2)
plt.axis('off')
plt.savefig('res.png')
frames.append(Image.fromarray(cv2.imread('res.png'), 'RGB'))
ind=ind+1
frame_one = frames[0]
frame_one.save("vis.gif", format="GIF", append_images=frames,save_all=True, duration=300, loop=0)
gif_image="vis.gif"
return outstr,gif_image
examples = [
['sample_test_images/59-11.png'],
['sample_test_images/59-21.png'],
['sample_test_images/59-32.png'],
['sample_test_images/59-37.png'],
['sample_test_images/91-47.png'],
['sample_test_images/91-49.png'],
]
def recognize_text(input_image):
x, x_mask=preprocess_img(input_image)
output_str, gifImage=test_error(x, x_mask)
return output_str,gifImage
title = "CALText Demo"
description = "<p style='text-align: center'>Gradio demo for an CALText model architecture <a href='https://github.com/nazar-khan/CALText'>[GitHub Code]</a> trained on the <a href='http://faculty.pucit.edu.pk/nazarkhan/work/urdu_ohtr/pucit_ohul_dataset.html'>PUCIT-OHUL</a> dataset. To use it, simply add your image, or click one of the examples to load them. </p>"
article = "<p style='text-align: center'></p>"
css = "#0 {object-fit: contain;} #1 {object-fit: contain;}"
inputs = gr.inputs.Image(label="Input Image")
demo = gr.Interface(fn=recognize_text,
inputs=inputs,
outputs=[gr.Textbox(label="Output"),
gr.Image(label="Attended Regions")],
examples=examples,
title=title,
description=description,
article=article,allow_flagging='never')
demo.launch()
|