mmagic.models.editors.inst_colorization.color_utils¶
Module Contents¶
Functions¶
|
Conversion images from xyz to rgb. |
|
Conversion images from lab to xyz. |
|
Conversion images from lab to rgb. |
|
Encode ab value into an index. |
|
Conversion images from rgb to xyz |
|
Conversion images from xyz to lab |
|
Conversion images from rgb to lab. |
|
Conversion images from rgb to lab. |
|
Add random color points sampled from ground truth based on: Number of |
- mmagic.models.editors.inst_colorization.color_utils.xyz2rgb(xyz)[source]¶
Conversion images from xyz to rgb.
- Parameters
xyz (tensor) – The images to be conversion
- Returns
The converted image
- Return type
out (tensor)
- mmagic.models.editors.inst_colorization.color_utils.lab2xyz(lab)[source]¶
Conversion images from lab to xyz.
- Parameters
lab (tensor) – The images to be conversion
- Returns
The converted image
- Return type
out (tensor)
- mmagic.models.editors.inst_colorization.color_utils.lab2rgb(lab_rs, color_data_opt)[source]¶
Conversion images from lab to rgb.
- Parameters
lab_rs (tensor) – The images to be conversion
color_data_opt (dict) – Config for image colorspace transformation. Include: l_norm, ab_norm, l_cent
- Returns
The converted image
- Return type
out (tensor)
- mmagic.models.editors.inst_colorization.color_utils.encode_ab_ind(data_ab, color_data_opt)[source]¶
Encode ab value into an index.
- Parameters
data_ab – Nx2xHxW from [-1,1]
color_data_opt – Config for image colorspace transformation. ab_max, ab_quant, ab_norm, ab_quant
- Returns
Nx1xHxW from [0,Q)
- mmagic.models.editors.inst_colorization.color_utils.rgb2xyz(rgb)[source]¶
Conversion images from rgb to xyz rgb from [0,1] xyz_from_rgb = np.array([[0.412453, 0.357580, 0.180423],
[0.212671, 0.715160, 0.072169], [0.019334, 0.119193, 0.950227]])
- Parameters
rgb (Tensor) – image in rgb colorspace
- Returns
image in xyz colorspace
- Return type
xyz (Tensor)
- mmagic.models.editors.inst_colorization.color_utils.xyz2lab(xyz)[source]¶
Conversion images from xyz to lab xyz from [0,1] factors: 0.95047, 1., 1.08883
- Parameters
xyz (Tensor) – image in xyz colorspace
- Returns
Image in lab colorspace
- Return type
out (Tensor)
- mmagic.models.editors.inst_colorization.color_utils.rgb2lab(rgb, color_opt)[source]¶
Conversion images from rgb to lab.
- Parameters
data_raw (tensor) – The images to be conversion
color_opt (dict) – Config for image colorspace transformation. Include: ab_thresh, ab_norm, sample_PS, mask_cent
- Returns
The converted image
- Return type
out (tensor)
- mmagic.models.editors.inst_colorization.color_utils.get_colorization_data(data_raw, color_opt, num_points=None)[source]¶
Conversion images from rgb to lab.
- Parameters
data_raw (tensor) – The images to be conversion
color_opt (dict) – Config for image colorspace transformation. Include: ab_thresh, ab_norm, sample_PS, mask_cent
- Returns
Output in add_color_patches_rand_gt
- Return type
results (dict)
- mmagic.models.editors.inst_colorization.color_utils.add_color_patches_rand_gt(data, color_opt, p=0.125, num_points=None, use_avg=True, samp='normal')[source]¶
Add random color points sampled from ground truth based on: Number of points.
- if num_points is 0, then sample from geometric distribution,
drawn from probability p
if num_points > 0, then sample that number of points
Location of points - if samp is ‘normal’, draw from N(0.5, 0.25) of image - otherwise, draw from U[0, 1] of image
- Parameters
data (tensor) – The images to be conversion
color_opt (dict) – Config for image colorspace transformation Include: ab_thresh, ab_norm, sample_PS, mask_cent
p (float) – Sampling geometric distribution, 1.0 means no hints
num_points (int) – Certain number of points
use_avg (bool) – Whether to use the mean when add color point Default: True.
samp (str) – Geometric distribution or uniform distribution when sample location. Default: normal.
- Returns
Result dict from :obj:
mmcv.BaseDataset.- Return type
results (dict)