0

i am trying to train text detection https://github.com/opencv/openvino_training_extensions/tree/develop/tensorflow_toolkit/text_detectio... and default it is set for image size 1280 * 768 , but i want to train it on cropped vehicle number plate, i have resized my images to 200*120px size with padding keeping the aspect ratio.

is there any doc available to understand config.yaml,

some fields there are like

min_area: 300

score_map_shape: [128,128]

train_image_shape : [512,512]

can someone please explain these. i tried setting train_image_shape with 200,120, and i got error operands could not be broadcast together with shapes (8,13,2)(8,14,2)

Thanks & Regards

Rawat

Rawat
  • 461
  • 3
  • 6
  • 23

1 Answers1

0

I suggest you refer to the config.py of the original implementation of PixelLink, available at the following link: https://github.com/ZJULearning/pixel_link/blob/master/config.py

Additionally, I would also encourage you to read the paper, “PixelLink: Detecting Scene Text via Instance Segmentation”, available at the following link: https://arxiv.org/abs/1801.01315

Rommel_Intel
  • 1,369
  • 1
  • 4
  • 8