Files
TexTeller/texteller/models/globals.py

24 lines
520 B
Python

# Formula image(grayscale) mean and variance
IMAGE_MEAN = 0.9545467
IMAGE_STD = 0.15394445
# Vocabulary size for TexTeller
VOCAB_SIZE = 15000
# Fixed size for input image for TexTeller
FIXED_IMG_SIZE = 448
# Image channel for TexTeller
IMG_CHANNELS = 1 # grayscale image
# Max size of token for embedding
MAX_TOKEN_SIZE = 1024
# Scaling ratio for random resizing when training
MAX_RESIZE_RATIO = 1.15
MIN_RESIZE_RATIO = 0.75
# Minimum height and width for input image for TexTeller
MIN_HEIGHT = 12
MIN_WIDTH = 30