[PYTHON] About HOG output of Scikit-Image

skimage.feature.hog (http://scikit-image.org/docs/dev/api/skimage.feature.html#hog)Extracts HOG features from the input image, but the output result is 1d-It is an array, and there is a problem that I do not know what it is. https://github.com/holtzhau/scikits.image/blob/master/skimage/feature/hog.You can read the code in py to see how to interpret this array.

### Important input arguments

 --```orientations```: Histogram bins (default = 9)
 --``` pixels_per_cell```: Cell size (default = (8,8))
 --``` cells_per_block```: Number of cells per block (default = (3, 3))

### Take a look at the output

 Line 180

#### **`hog.py`**

return normalised_blocks.ravel()

I wanted you to return it without raveling here. Looking at the declaration of normalised_blocks,

Line 160


n_blocksx = (n_cellsx - bx) + 1
n_blocksy = (n_cellsy - by) + 1
normalised_blocks = np.zeros((n_blocksx, n_blocksy, bx, by, orientations))

The variables used here are declared around line 100


sx, sy = image.shape
cx, cy = pixels_per_cell
bx, by = cells_per_block

n_cellsx = int(np.floor(sx // cx))  # number of cells in x
n_cellsy = int(np.floor(sy // cy))  # number of cells in y

So, the 1-d array (let's say `` `retval```) obtained by turning hog is

retval.reshape((n_blocksx, n_blocksy, bx, by, orientations))

If you give it as, you can return it to its original shape.

Usage example

For example, the maximum gradient direction for each block can be visualized.


from skimage.feature import hog
from skimage.data import camera
import matplotlib.pyplot as plt

img = camera()
orientations = 9
pixels_per_cell = (8, 8)
cells_per_block = (3, 3)

h = hog(img, orientations=orientations, pixels_per_cell=pixels_per_cell, cells_per_block=cells_per_block)

sx, sy = img.shape[:2]
cx, cy = (8, 8)
bx, by = (3, 3)
n_cellsx = int(np.floor(sx // cx))  # number of cells in x
n_cellsy = int(np.floor(sy // cy))  # number of cells in y
n_blocksx = (n_cellsx - bx) + 1
n_blocksy = (n_cellsy - by) + 1

himg = h.reshape((n_blocksx * n_blocksy, bx, by, orientations))
vis = np.array([np.argmax(x.sum(axis=(0, 1))) for x in himg]).reshape((n_blocksx, n_blocksy))

plt.subplot(1, 2, 1)
plt.subplot(1, 2, 2)
plt.imshow(vis, interpolation='nearest')


Recommended Posts

About HOG output of Scikit-Image
About all of numpy
About assignment of numpy.ndarray
About MultiIndex of pandas
About variable of chainer
Basics of python: Output
About max_iter of LogisticRegression () of scikit-learn
About Japanese path of pyminizip
About the ease of Python
About Japanese support of cometchat
About all of numpy (2nd)
About cost calculation of MeCab
About approximate fractions of pi
About the components of Luigi
Debug output of chalice command
About the features of Python
About data management of anvil-app-server
About the return value of pthread_mutex_init ()
About the return value of the histogram.
About the basic type of Go
About the upper limit of threads-max
About circular crossover of genetic algorithms
About the behavior of yield_per of SqlAlchemy
About import error of PyQt5.QtWidgets (Anaconda)
About the size of matplotlib points
About color halftone processing of images
About the basics list of Python basics