Differential Binarization model #2095
Conversation
| output = self.image_converter( | ||
| { | ||
| "images": x, | ||
| "bounding_boxes": y, |
There was a problem hiding this comment.
I don't think we need any bounding boxes for this task.
sachinprasadhs
added
the
WIP
| return (self.x, self.y) | ||
|
|
||
|
|
||
| def shrink_polygan(polygon, offset): |
| class Point: | ||
| def __init__(self, x, y): | ||
| self.x = x | ||
| self.y = y | ||
|
|
||
| def __add__(self, other): | ||
| return Point(self.x + other.x, self.y + other.y) | ||
|
|
||
| def __sub__(self, other): | ||
| return Point(self.x - other.x, self.y - other.y) | ||
|
|
||
| def __neg__(self): | ||
| return Point(-self.x, -self.y) | ||
|
|
||
| def cross(self, other): | ||
| return self.x * other.y - self.y * other.x | ||
|
|
||
| def to_tuple(self): | ||
| return (self.x, self.y) | ||
|
|
There was a problem hiding this comment.
Is there any other way you can try to achive this better without using class?
|
|
||
|
|
||
| # cv2.fillpoly function with keras.ops | ||
| def fill_poly_keras(vertices, image_shape): |
There was a problem hiding this comment.
you can just mention fill_poly and remove the cv2 mention here and in he description and description of what this function does.
| return int(height) if height >= 0.1 else 0 | ||
|
|
||
|
|
||
| # project point to line |
There was a problem hiding this comment.
Remove all these comments, instead make the description better
| high = mid | ||
|
|
||
| height = (low + high) / 2 | ||
| height = (low + high) / 2 |
|
|
||
| # get line of height | ||
| def get_line_height(poly): | ||
| return binary_search_smallest_width(poly) |
There was a problem hiding this comment.
You can avoid this funtion and instead call binary_search_smallest_width directly
| """ | ||
| Shrinks a polygon inward by moving each point toward the center. |
There was a problem hiding this comment.
All the docstring text should immediately follow """, change it to, and this style should be applied everywhere
"""Shrinks a polygon inward by moving each point toward the center.
| @preprocessing_function | ||
| def generate_postprocess(self,x): | ||
| ''' | ||
| Generates postprocess function to convert probability map of | ||
| model output to polygon | ||
| ''' | ||
| probability_maps,threshold_maps = x["probability_maps"], x["threshold_maps"] | ||
| binary_maps = 1.0 / (1.0 + keras.ops.exp(-50.0 * (probability_maps - threshold_maps))) | ||
| outputs = keras.layers.Concatenate(axis=-1)( | ||
| [probability_maps, threshold_maps, binary_maps]) | ||
| return outputs | ||
|
|
There was a problem hiding this comment.
why do we need this? we don't have .generate function for diff_bin right?
| target_size=(640, 640), | ||
| shrink_ratio=0.3, |
There was a problem hiding this comment.
target_size will be image_size and will be handled in ImageConverter, not in this.
Where are we using this shrink_ratio here.
Differential Binarization model.