Skip to content

S
SSD_landmarks
Commit 77ba43b1 authored by Trái Vú Sữa's avatar Trái Vú Sữa 😄
Browse files
Options

fix #2

parent 186646b8
Show whitespace changes
Inline Side-by-side
Showing with 109 additions and 12 deletions
datasets/data_augment.py
View file @ 77ba43b1
...@@ -211,6 +211,7 @@ def _resize_subtract_mean(image, insize, rgb_mean): ...@@ -211,6 +211,7 @@ def _resize_subtract_mean(image, insize, rgb_mean):
image = cv2.resize(image, (insize, insize), interpolation=interp_method) image = cv2.resize(image, (insize, insize), interpolation=interp_method)
image = image.astype(np.float32) image = image.astype(np.float32)
image -= rgb_mean image -= rgb_mean
image /= 128.0
return image.transpose(2, 0, 1) return image.transpose(2, 0, 1)
......
datasets/wider_face.py
View file @ 77ba43b1
import os import os
import os.path import os.path
import sys
import torch import torch
import torch.utils.data as data import torch.utils.data as data
import cv2 import cv2
import numpy as np import numpy as np
import torch.nn.functional as F from imgaug import augmenters as iaa
import skimage.transform
import torchvision.transforms as transforms
class FaceDataset(data.Dataset): class FaceDataset(data.Dataset):
...@@ -15,15 +12,15 @@ class FaceDataset(data.Dataset): ...@@ -15,15 +12,15 @@ class FaceDataset(data.Dataset):
super(FaceDataset, self).__init__() super(FaceDataset, self).__init__()
self.path_images, self.labels = self.read_file(root_path, file_name) self.path_images, self.labels = self.read_file(root_path, file_name)
self.preproc = preproc self.preproc = preproc
self.augment = ImgAugTransform()
self.target_transform = target_transform self.target_transform = target_transform
def __len__(self): def __len__(self):
return len(self.path_images) return len(self.path_images)
def __getitem__(self, idx): def __getitem__(self, idx):
img = cv2.imread(self.path_images[idx]) img_raw = cv2.imread(self.path_images[idx])
img = self.augment(img_raw)
labels = self.labels[idx] labels = self.labels[idx]
annotations = np.zeros((0, 15)) annotations = np.zeros((0, 15))
...@@ -55,7 +52,7 @@ class FaceDataset(data.Dataset): ...@@ -55,7 +52,7 @@ class FaceDataset(data.Dataset):
annotations = np.append(annotations, annotation, axis=0) annotations = np.append(annotations, annotation, axis=0)
target = np.array(annotations) target = np.array(annotations)
debug = False debug = True
if debug: if debug:
img_debug = img.copy() img_debug = img.copy()
for index, b in enumerate(annotations): for index, b in enumerate(annotations):
...@@ -69,8 +66,9 @@ class FaceDataset(data.Dataset): ...@@ -69,8 +66,9 @@ class FaceDataset(data.Dataset):
cv2.circle(img_debug, (b[10], b[11]), 1, (0, 255, 0), 4) cv2.circle(img_debug, (b[10], b[11]), 1, (0, 255, 0), 4)
cv2.circle(img_debug, (b[12], b[13]), 1, (255, 0, 0), 4) cv2.circle(img_debug, (b[12], b[13]), 1, (255, 0, 0), 4)
name = "test_data.jpg" img_show = np.hstack((img_debug, img_raw))
cv2.imwrite(name, img_debug) cv2.imshow("test", img_show)
cv2.waitKey()
if self.preproc is not None: if self.preproc is not None:
img, target = self.preproc(img, target) img, target = self.preproc(img, target)
...@@ -78,8 +76,8 @@ class FaceDataset(data.Dataset): ...@@ -78,8 +76,8 @@ class FaceDataset(data.Dataset):
labels = target[:, -1] labels = target[:, -1]
landms = target[:, 4:14] landms = target[:, 4:14]
# TODO write landms to target_transforms # TODO write landms to target_transforms
if self.target_transform: # if self.target_transform:
boxes,landms, labels = self.target_transform(boxes,landms,labels) # boxes,landms, labels = self.target_transform(boxes,landms,labels)
return torch.from_numpy(img), target return torch.from_numpy(img), target
@staticmethod @staticmethod
...@@ -105,3 +103,101 @@ class FaceDataset(data.Dataset): ...@@ -105,3 +103,101 @@ class FaceDataset(data.Dataset):
labels.append(label) labels.append(label)
words.append(labels) words.append(labels)
return path_images, words return path_images, words
class ImgAugTransform:
def __init__(self):
self.aug = iaa.Sequential([
iaa.Sometimes(0.15, iaa.Grayscale(alpha=(0.05, 1.0))),
iaa.Sometimes(0.2,
iaa.OneOf([
iaa.Sometimes(0.25, iaa.MotionBlur(k=5, angle=[-45, 45])),
iaa.Sometimes(0.25, iaa.GaussianBlur(sigma=(0, 1.3)))
])),
iaa.Sometimes(0.2,
iaa.OneOf([
iaa.Sharpen(alpha=(0.0, 1.0), lightness=(0.75, 2.0)),
iaa.Emboss(alpha=(0.0, 1.0), strength=(0.5, 1.5)),
iaa.EdgeDetect(alpha=(0.0, 1.0)),
iaa.DirectedEdgeDetect(alpha=(0.0, 1.0), direction=(0.0, 1.0)),
iaa.Canny(
alpha=(0.0, 0.5),
colorizer=iaa.RandomColorsBinaryImageColorizer(
color_true=255,
color_false=0
)
)
])),
iaa.Sometimes(0.2,
iaa.OneOf([
iaa.GammaContrast((0.5, 2.0)),
iaa.GammaContrast((0.5, 2.0), per_channel=True),
iaa.SigmoidContrast(gain=(3, 10), cutoff=(0.4, 0.6)),
iaa.SigmoidContrast(gain=(3, 10), cutoff=(0.4, 0.6), per_channel=True),
iaa.LogContrast(gain=(0.6, 1.4)),
iaa.LinearContrast((0.4, 1.6)),
iaa.AllChannelsCLAHE(clip_limit=(1, 10)),
iaa.CLAHE(clip_limit=(1, 10)),
iaa.Alpha((0.0, 1.0), iaa.AllChannelsHistogramEqualization()),
iaa.Alpha((0.0, 1.0), iaa.HistogramEqualization()),
iaa.pillike.Autocontrast(cutoff=(0, 15.0)),
])),
iaa.Sometimes(0.2,
iaa.OneOf([
iaa.MultiplyBrightness(),
iaa.Add((-40, 40)),
iaa.Multiply((0.5, 1.5)),
iaa.Multiply((0.5, 1.5), per_channel=0.5),
iaa.Sometimes(0.15, iaa.MultiplyAndAddToBrightness(mul=(0.5, 1.5), add=(-30, 30)))
])),
iaa.Sometimes(0.15,
iaa.OneOf([iaa.Dropout(p=(0, 0.1)),
iaa.Dropout2d(p=0.1),
iaa.SaltAndPepper(0.1),
])),
iaa.Sometimes(0.25,
iaa.OneOf([iaa.BlendAlphaElementwise((0, 1.0), iaa.AddToHue(100)),
iaa.BlendAlphaSimplexNoise(iaa.EdgeDetect(1.0), upscale_method="linear"),
iaa.BlendAlphaElementwise([0.25, 0.75], iaa.MedianBlur(5)),
iaa.BlendAlphaSomeColors(iaa.AveragePooling(7), from_colorspace="BGR"),
iaa.BlendAlphaHorizontalLinearGradient(
iaa.TotalDropout(1.0),
min_value=0.2, max_value=0.8),
iaa.BlendAlphaHorizontalLinearGradient(
iaa.AveragePooling(11),
start_at=(0.0, 1.0), end_at=(0.0, 1.0)),
])),
iaa.Sometimes(0.15,
iaa.OneOf([
iaa.Clouds(),
iaa.Fog(),
iaa.Snowflakes(flake_size=(0.1, 0.4), speed=(0.01, 0.05)),
iaa.Rain(speed=(0.1, 0.3))
])),
# iaa.Sometimes(0.15,
# iaa.OneOf([
# iaa.Cartoon(),
# iaa.Cartoon(blur_ksize=3, segmentation_size=1.0,
# saturation=2.0, edge_prevalence=1.0),
# iaa.Superpixels(p_replace=0.5, n_segments=64),
# iaa.Superpixels(p_replace=(0.1, 1.0), n_segments=(16, 128)),
# iaa.UniformVoronoi(250, p_replace=0.9, max_size=None),
# ])),
], random_order=True)
def __call__(self, img):
img = np.array(img)
return self.aug.augment_image(img)
if __name__ == "__main__":
from datasets.data_augment import preproc
loader = FaceDataset(root_path=os.path.join('/home/can/AI_Camera/EfficientFaceNet/data/widerface/train/images'), file_name='label.txt',
preproc=preproc(300, (127, 127, 127)))
print(len(loader))
for i in range(0, len(loader)):
print("\n****")
print(i)
a = loader.__getitem__(i)
\ No newline at end of file
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment