def pdf2latex(model, img_path):
buckets = [[240, 100], [320, 80], [400, 80], [400, 100], [480, 80],
[480, 100], [560, 80], [560, 100], [640, 80], [640, 100],
[720, 80], [720, 100], [720, 120], [720, 200], [800, 100],
[800, 320], [1000, 200], [1000, 400], [1200, 200], [1600, 200],
[1600, 1600]]
dir_output = "tmp/"
name = img_path.split('/')[-1].split('.')[0]
run(
"magick convert -density {} -quality {} {} {}".format(
200, 100, img_path, dir_output + "{}.png".format(name)), TIMEOUT)
img_path = dir_output + "{}.png".format(name)
crop_image(img_path, img_path)
pad_image(img_path, img_path, buckets=buckets)
downsample_image(img_path, img_path, 2)
img = imread(img_path)
img = greyscale(img)
hyps = model.predict(img)
# model.logger.info(hyps[0])
return hyps[0], img_path
def img2latex(model, img_path):
if not os.path.exists(img_path):
print('image not exists')
if img_path[-3:] == "png":
img = imread(img_path)
elif img_path[-3:] == "pdf":
# call magick to convert the pdf into a png file
buckets = [[240, 100], [320, 80], [400, 80], [400, 100], [480, 80],
[480, 100], [560, 80], [560, 100], [640, 80], [640, 100],
[720, 80], [720, 100], [720, 120], [720, 200], [800, 100],
[800, 320], [1000, 200], [1000, 400], [1200, 200],
[1600, 200], [1600, 1600]]
dir_output = "tmp/"
name = img_path.split('/')[-1].split('.')[0]
run(
"magick convert -density {} -quality {} {} {}".format(
200, 100, img_path, dir_output + "{}.png".format(name)),
TIMEOUT)
img_path = dir_output + "{}.png".format(name)
crop_image(img_path, img_path)
pad_image(img_path, img_path, buckets=buckets)
downsample_image(img_path, img_path, 2)
img = imread(img_path)
img = greyscale(img)
hyps = model.predict(img)
return hyps[0]
def interactive_shell(model):
"""Creates interactive shell to play with model
"""
model.logger.info("""
This is an interactive mode.
To exit, enter 'exit'.
Enter a path to a file
input> data/images_test/0.png""")
while True:
try:
# for python 2
img_path = raw_input("input> ")
except NameError:
# for python 3
img_path = input("input> ")
if img_path == "exit":
break
dir_output = "/tmp/"
if img_path[-3:] == "png":
img = Image.open(img_path)
img = img.resize((80, 100), Image.BILINEAR)
img.show()
img = numpy.array(img)
# crop_image(img_path, tmp_img)
# pad_image(tmp_img, tmp_img, buckets=None)
# downsample_image(tmp_img, tmp_img, 2)
# img = imread(tmp_img)
elif img_path[-3:] == "pdf":
# call magick to convert the pdf into a png file
buckets = [[240, 100], [320, 80], [400, 80], [400, 100], [480, 80],
[480, 100],
[560, 80], [560, 100], [640, 80], [640, 100], [720, 80],
[720, 100], [720, 120], [720, 200], [800, 100],
[800, 320], [1000, 200], [1000, 400], [1200, 200],
[1600, 200], [1600, 1600]]
name = img_path.split('/')[-1].split('.')[0]
run(
"magick convert -density {} -quality {} {} {}".format(
200, 100, img_path, dir_output + "{}.png".format(name)),
TIMEOUT)
img_path = dir_output + "{}.png".format(name)
crop_image(img_path, img_path)
pad_image(img_path, img_path, buckets=buckets)
downsample_image(img_path, img_path, 2)
img = imread(img_path)
img = greyscale(img)
hyps = model.predict(img)
model.logger.info(hyps[0])
def img2latex(model,
img_path,
downsample_image_ratio=1,
cropping=False,
padding=False,
img_augment=None,
gray_scale=True):
dir_output = "tmp/"
run(['mkdir -p tmp'], TIMEOUT)
name = img_path.split('/')[-1].split('.')[0]
buckets = [[240, 100], [320, 80], [400, 80], [400, 100], [480, 80],
[480, 100], [560, 80], [560, 100], [640, 80], [640, 100],
[720, 80], [720, 100], [720, 120], [720, 200], [800, 100],
[800, 320], [1000, 200], [1000, 400], [1200, 200], [1600, 200],
[1600, 1600]]
img_path_tmp = dir_output + "{}.png".format(name)
if cropping:
crop_image(img_path, img_path_tmp)
if padding:
pad_image(img_path_tmp if cropping else img_path,
img_path_tmp,
buckets=buckets)
if downsample_image_ratio != 1:
if cropping or padding:
downsample_image(img_path_tmp,
img_path_tmp,
ratio=downsample_image_ratio)
else:
downsample_image(img_path,
img_path_tmp,
ratio=downsample_image_ratio)
if cropping or padding or downsample_image_ratio != 1:
img = imread(img_path_tmp)
else:
img = imread(img_path)
if img_augment:
img = img_augment.augment_image(img)
img_obj = Image.fromarray(img)
img_obj.save(img_path_tmp)
if gray_scale:
last = greyscale(img)
else:
last = img
hyps = model.predict(last)
return hyps[0], img, os.path.abspath(img_path_tmp)
def pdf2latex(model, pdf_path):
"""
Make prediction for PDF
:param model: model to be used
:param pdf_path: PDF location
:return:
"""
buckets = [
[240, 100],
[320, 80],
[400, 80],
[400, 100],
[480, 80],
[480, 100],
[560, 80],
[560, 100],
[640, 80],
[640, 100],
[720, 80],
[720, 100],
[720, 120],
[720, 200],
[800, 100],
[800, 320],
[1000, 200],
[1000, 400],
[1200, 200],
[1600, 200],
[1600, 1600],
]
dir_output = "tmp/"
name = pdf_path.split("/")[-1].split(".")[0]
run(
"magick convert -density {} -quality {} {} {}".format(
200, 100, pdf_path, dir_output + "{}.png".format(name)
),
TIMEOUT,
)
pdf_path = dir_output + "{}.png".format(name)
crop_image(pdf_path, pdf_path)
pad_image(pdf_path, pdf_path, buckets=buckets)
downsample_image(pdf_path, pdf_path, 2)
img = imread(pdf_path)
img = greyscale(img)
hyps = model.predict(img)
# model.logger.info(hyps[0])
return hyps[0], pdf_path
def index():
if request.method == 'GET':
return render_template('index.html')
if request.method == 'POST':
file = request.files['image']
filename = secure_filename(file.filename)
savedpath = os.path.join(UPLOAD_FOLDER, filename)
file.save(savedpath)
print("New uploaded file: [%s]" % savedpath)
try:
if filename[-3:] == "png":
img = imread(savedpath)
elif filename[-3:] == "pdf":
# call magick to convert the pdf into a png file
buckets = [
[240, 100], [320, 80], [400, 80], [400, 100], [480, 80], [480, 100],
[560, 80], [560, 100], [640, 80], [640, 100], [720, 80], [720, 100],
[720, 120], [720, 200], [800, 100], [800, 320], [1000, 200],
[1000, 400], [1200, 200], [1600, 200], [1600, 1600]
]
dir_output = "tmp/"
name = savedpath.split('/')[-1].split('.')[0]
run("magick convert -density {} -quality {} {} {}".format(200, 100,
savedpath, dir_output+"{}.png".format(name)), TIMEOUT)
savedpath = dir_output + "{}.png".format(name)
crop_image(savedpath, savedpath)
pad_image(savedpath, savedpath, buckets=buckets)
downsample_image(savedpath, savedpath, 2)
img = imread(savedpath)
else:
raise "Unsupported file format"
img = greyscale(img)
hyps = model.predict(img)
model.logger.info(hyps[0])
message = hyps[0]
except Exception as e:
print(e)
message = "Error loading. Try a different image"
return render_template('index.html', latexstr=message, filename=file.filename)
return "NOT Support Method"
def interactive_shell(model):
"""Creates interactive shell to play with model
"""
model.logger.info("""
This is an interactive mode.
To exit, enter 'exit'.
Enter a path to a file
input> data/images_test/0.png""")
while True:
try:
# for python 2
img_path = raw_input("input> ")
except NameError:
# for python 3
img_path = input("input> ")
if img_path == "exit":
break
if not os.path.exists(img_path):
print('image not exists')
continue
if img_path[-3:] == "png":
img = imread(img_path)
elif img_path[-3:] == "pdf":
# call magick to convert the pdf into a png file
buckets = [[240, 100], [320, 80], [400, 80], [400, 100], [480, 80],
[480, 100],
[560, 80], [560, 100], [640, 80], [640, 100], [720, 80],
[720, 100], [720, 120], [720, 200], [800, 100],
[800, 320], [1000, 200], [1000, 400], [1200, 200],
[1600, 200], [1600, 1600]]
dir_output = "tmp/"
name = img_path.split('/')[-1].split('.')[0]
run(
"magick convert -density {} -quality {} {} {}".format(
200, 100, img_path, dir_output + "{}.png".format(name)),
TIMEOUT)
img_path = dir_output + "{}.png".format(name)
crop_image(img_path, img_path)
pad_image(img_path, img_path, buckets=buckets)
downsample_image(img_path, img_path, 2)
img = imread(img_path)
img = greyscale(img)
hyps = model.predict(img)
model.logger.info(hyps[0])
def img2latex(
model,
img,
downsample_image_ratio=1,
cropping=False,
padding=False,
img_augment=None,
gray_scale=True,
):
"""
Predict a latex code for an input equation image.
:param model: model to be used
:param img: input equation image
:param downsample_image_ratio: down sampling ratio
:param cropping: whether to crop
:param padding: whether to pad
:param img_augment: img augmentation filter
:param gray_scale:whether to gray scale
:return: latex prediction, processed img, processed img location
"""
dir_output = "tmp/"
run(["mkdir -p tmp"], TIMEOUT)
name = str(uuid.uuid4())
img_path = os.path.join("tmp/", f"{name}.png")
img.save(img_path)
buckets = [
[240, 100],
[320, 80],
[400, 80],
[400, 100],
[480, 80],
[480, 100],
[560, 80],
[560, 100],
[640, 80],
[640, 100],
[720, 80],
[720, 100],
[720, 120],
[720, 200],
[800, 100],
[800, 320],
[1000, 200],
[1000, 400],
[1200, 200],
[1600, 200],
[1600, 1600],
]
img_path_tmp = dir_output + "{}.png".format(name)
if cropping:
crop_image(img_path, img_path_tmp)
if padding:
pad_image(img_path_tmp if cropping else img_path, img_path_tmp, buckets=buckets)
if downsample_image_ratio != 1:
if cropping or padding:
downsample_image(img_path_tmp, img_path_tmp, ratio=downsample_image_ratio)
else:
downsample_image(img_path, img_path_tmp, ratio=downsample_image_ratio)
if cropping or padding or downsample_image_ratio != 1:
img = imread(img_path_tmp)
else:
img = imread(img_path)
if img_augment:
img = img_augment.augment_image(img)
img_obj = Image.fromarray(img)
img_obj.save(img_path_tmp)
if gray_scale:
last = greyscale(img)
else:
last = img
hyps = model.predict(last)
return hyps[0], img, os.path.abspath(img_path_tmp)
