updated notebooks

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

import os

import os

import sys

import sys

sys.path.insert(0, os.path.abspath('..'))

sys.path.insert(0, os.path.abspath('..'))

from autocnet.examples import get_path

from autocnet.examples import get_path

from autocnet.graph.network import CandidateGraph

from autocnet.graph.network import CandidateGraph

from autocnet.graph.edge import Edge

from autocnet.graph.edge import Edge

from autocnet.matcher.matcher import FlannMatcher

from autocnet.matcher.feature import FlannMatcher

from autocnet.matcher import ciratefi

from autocnet.matcher import ciratefi

from autocnet.matcher import subpixel as sp

from autocnet.matcher import subpixel as sp

from scipy.misc import imresize

from scipy.misc import imresize

import math

import math

import warnings

import warnings

import cv2

import cv2

from bisect import bisect_left

from bisect import bisect_left

from scipy.ndimage.interpolation import rotate

from scipy.ndimage.interpolation import rotate

from IPython.display import display

from IPython.display import display

warnings.filterwarnings('ignore')

warnings.filterwarnings('ignore')

%matplotlib inline

%matplotlib inline

%pylab inline

%pylab inline

```

```

%% Output

%% Output

    Populating the interactive namespace from numpy and matplotlib

    Populating the interactive namespace from numpy and matplotlib

    WARNING: pylab import has clobbered these variables: ['e']

    WARNING: pylab import has clobbered these variables: ['e']

    `%matplotlib` prevents importing * from pylab and numpy

    `%matplotlib` prevents importing * from pylab and numpy

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

# Create Basic Structures

# Create Basic Structures

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#Point to the adjacency Graph

#Point to the adjacency Graph

adjacency = get_path('three_image_adjacency.json')

adjacency = get_path('three_image_adjacency.json')

basepath = get_path('Apollo15')

basepath = get_path('Apollo15')

cg = CandidateGraph.from_adjacency(adjacency, basepath=basepath)

cg = CandidateGraph.from_adjacency(adjacency, basepath=basepath)

#Apply SIFT to extract features

#Apply SIFT to extract features

cg.extract_features(method='sift', extractor_parameters={'nfeatures':300})

cg.extract_features(method='sift', extractor_parameters={'nfeatures':300})

#Match

#Match

cg.match_features()

cg.match_features()

# Perform the symmetry check

# Perform the symmetry check

cg.symmetry_checks()

cg.symmetry_checks()

# Perform the ratio check

# Perform the ratio check

cg.ratio_checks(clean_keys = ['symmetry'])

cg.ratio_checks(clean_keys = ['symmetry'])

# Create fundamental matrix

# Create fundamental matrix

cg.compute_fundamental_matrices(clean_keys = ['symmetry', 'ratio'])

cg.compute_fundamental_matrices(clean_keys = ['symmetry', 'ratio'])

# Step: Compute the homographies and apply RANSAC

# Step: Compute the homographies and apply RANSAC

cg.compute_homographies(clean_keys=['symmetry', 'ratio'])

cg.compute_homographies(clean_keys=['symmetry', 'ratio'])

# Step: Compute subpixel offsets for candidate points

# Step: Compute subpixel offsets for candidate points

cg.subpixel_register(clean_keys=['ransac'])

cg.subpixel_register(clean_keys=['ransac'])

cg.suppress(clean_keys=['symmetry', 'ratio', 'subpixel'])

cg.suppress(clean_keys=['symmetry', 'ratio', 'subpixel'])

```

```

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

# Do Stuff

# Do Stuff

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

from scipy.ndimage.interpolation import zoom

from scipy.ndimage.interpolation import zoom

from scipy.stats.stats import pearsonr

from scipy.stats.stats import pearsonr

figsize(10,10)

figsize(10,10)

e = cg.edge[0][2]

e = cg.edge[0][2]

matches = e.matches

matches = e.matches

clean_keys = ['subpixel']

clean_keys = ['subpixel']

full_offsets = np.zeros((len(matches), 3))

full_offsets = np.zeros((len(matches), 3))

if clean_keys:

if clean_keys:

    matches, mask = e._clean(clean_keys)

    matches, mask = e._clean(clean_keys)

# Preallocate the numpy array to avoid appending and type conversion

# Preallocate the numpy array to avoid appending and type conversion

edge_offsets = np.empty((len(matches),3))

edge_offsets = np.empty((len(matches),3))

# for each edge, calculate this for each keypoint pair

# for each edge, calculate this for each keypoint pair

for i, (idx, row) in enumerate(matches.iterrows()):

for i, (idx, row) in enumerate(matches.iterrows()):

    s_idx = int(row['source_idx'])

    s_idx = int(row['source_idx'])

    d_idx = int(row['destination_idx'])

    d_idx = int(row['destination_idx'])

    s_kps = e.source.get_keypoints().iloc[s_idx]

    s_kps = e.source.get_keypoints().iloc[s_idx]

    d_kps = e.destination.get_keypoints().iloc[d_idx]

    d_kps = e.destination.get_keypoints().iloc[d_idx]

    s_keypoint = e.source.get_keypoints().iloc[s_idx][['x', 'y']].values

    s_keypoint = e.source.get_keypoints().iloc[s_idx][['x', 'y']].values

    d_keypoint = e.destination.get_keypoints().iloc[d_idx][['x', 'y']].values

    d_keypoint = e.destination.get_keypoints().iloc[d_idx][['x', 'y']].values

    # Get the template and search windows

    # Get the template and search windows

    s_template = sp.clip_roi(e.source.geodata, s_keypoint, 5)

    s_template = sp.clip_roi(e.source.geodata, s_keypoint, 5)

    s_template = rotate(s_template, 0)

    s_template = rotate(s_template, 0)

    s_template = imresize(s_template, 1.)

    s_template = imresize(s_template, 1.)

    d_search = sp.clip_roi(e.destination.geodata, d_keypoint, 11)

    d_search = sp.clip_roi(e.destination.geodata, d_keypoint, 11)

    d_search = rotate(d_search, 0)

    d_search = rotate(d_search, 0)

    d_search = imresize(d_search, 1.)

    d_search = imresize(d_search, 1.)

    imshow(s_template, cmap='Greys')

    imshow(s_template, cmap='Greys')

    show()

    show()

    imshow(d_search, cmap='Greys')

    imshow(d_search, cmap='Greys')

    show()

    show()

    result = ciratefi.ciratefi(s_template, d_search, upsampling=10., alpha=math.pi/4,

    result = ciratefi.ciratefi(s_template, d_search, upsampling=10., alpha=math.pi/4,

                     cifi_thresh=70, rafi_thresh=70, tefi_thresh=100,

                     cifi_thresh=70, rafi_thresh=70, tefi_thresh=100,

                     use_percentile=True, radii=list(range(1,3)), verbose=True)

                     use_percentile=True, radii=list(range(1,3)), verbose=True)

    print(result)

    print(result)

    break

    break

```

```

%% Output

%% Output

    ---------------------------------------------------------------------------

    ---------------------------------------------------------------------------

    ValueError                                Traceback (most recent call last)

    ValueError                                Traceback (most recent call last)

    <ipython-input-10-52559c337e48> in <module>()

    <ipython-input-10-52559c337e48> in <module>()

         41     result = ciratefi.ciratefi(s_template, d_search, upsampling=10., alpha=math.pi/4,

         41     result = ciratefi.ciratefi(s_template, d_search, upsampling=10., alpha=math.pi/4,

         42                      cifi_thresh=70, rafi_thresh=70, tefi_thresh=100,

         42                      cifi_thresh=70, rafi_thresh=70, tefi_thresh=100,

    ---> 43                      use_percentile=True, radii=list(range(1,3)), verbose=True)

    ---> 43                      use_percentile=True, radii=list(range(1,3)), verbose=True)

         44     print(result)

         44     print(result)

         45     break

         45     break

    /Users/kelvinrodriguez/Repos/autocnet/autocnet/matcher/ciratefi.py in ciratefi(template, search_image, upsampling, cifi_thresh, rafi_thresh, tefi_thresh, use_percentile, alpha, scales, radii, verbose)

    /Users/kelvinrodriguez/Repos/autocnet/autocnet/matcher/ciratefi.py in ciratefi(template, search_image, upsampling, cifi_thresh, rafi_thresh, tefi_thresh, use_percentile, alpha, scales, radii, verbose)

        613     # Perform last filter tefi

        613     # Perform last filter tefi

        614     results = tefi(template, search_image, sg_candidate_points, best_scales, best_rotation,

        614     results = tefi(template, search_image, sg_candidate_points, best_scales, best_rotation,

    --> 615                    thresh=tefi_thresh, alpha=math.pi/4, use_percentile=True, upsampling=upsampling, verbose=verbose)

    --> 615                    thresh=tefi_thresh, alpha=math.pi/4, use_percentile=True, upsampling=upsampling, verbose=verbose)

        616

        616

        617     # return the points found

        617     # return the points found

    /Users/kelvinrodriguez/Repos/autocnet/autocnet/matcher/ciratefi.py in tefi(template, search_image, candidate_pixels, best_scales, best_angles, scales, upsampling, thresh, alpha, use_percentile, verbose)

    /Users/kelvinrodriguez/Repos/autocnet/autocnet/matcher/ciratefi.py in tefi(template, search_image, candidate_pixels, best_scales, best_angles, scales, upsampling, thresh, alpha, use_percentile, verbose)

        460     best_angles = np.asarray(best_angles, dtype=np.float32)

        460     best_angles = np.asarray(best_angles, dtype=np.float32)

        461     if not best_angles.size or not np.any(best_angles):

        461     if not best_angles.size or not np.any(best_angles):

    --> 462         raise ValueError('Input best angle list is empty')

    --> 462         raise ValueError('Input best angle list is empty')

        463

        463

        464     best_scales = np.asarray(best_scales, dtype=float)

        464     best_scales = np.asarray(best_scales, dtype=float)

    ValueError: Input best angle list is empty

    ValueError: Input best angle list is empty

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

```

```

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

import os

import os

import sys

import sys

sys.path.insert(0, os.path.abspath('..'))

sys.path.insert(0, os.path.abspath('..'))

from autocnet.examples import get_path

from autocnet.examples import get_path

from autocnet.graph.network import CandidateGraph

from autocnet.graph.network import CandidateGraph

from autocnet.matcher.matcher import FlannMatcher

from autocnet.matcher.feature import FlannMatcher

from autocnet.matcher.suppression_funcs import distance

from autocnet.matcher.suppression_funcs import distance

from IPython.display import display

from IPython.display import display

%pylab inline

%pylab inline

```

```

%% Output

%% Output

    Populating the interactive namespace from numpy and matplotlib

    Populating the interactive namespace from numpy and matplotlib

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#Point to the adjacency Graph

#Point to the adjacency Graph

adjacency = get_path('two_image_adjacency.json')

adjacency = get_path('two_image_adjacency.json')

basepath = get_path('Apollo15')

basepath = get_path('Apollo15')

cg = CandidateGraph.from_adjacency(adjacency, basepath=basepath)

cg = CandidateGraph.from_adjacency(adjacency, basepath=basepath)

#Apply SIFT to extract features

#Apply SIFT to extract features

cg.extract_features(method='sift')

cg.extract_features(method='sift')

#Match

#Match

cg.match_features()

cg.match_features()

#Apply outlier detection

#Apply outlier detection

cg.symmetry_checks()

cg.symmetry_checks()

cg.ratio_checks()

cg.ratio_checks()

m = cg.edge[0][1].masks

m = cg.edge[0][1].masks

#Compute a fundamental matrix

#Compute a fundamental matrix

cg.compute_fundamental_matrices(clean_keys=['ratio', 'symmetry'])

cg.compute_fundamental_matrices(clean_keys=['ratio', 'symmetry'])

```

```

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

## Suppression

## Suppression

Create a suppression object using a default error tolerance and count.  Supply a custom function that suppresses based upon the distance between matches.

Create a suppression object using a default error tolerance and count.  Supply a custom function that suppresses based upon the distance between matches.

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#figsize(12,12)

#figsize(12,12)

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance)

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance)

# Plot, in blue the points that passed all outlier detectors so far

# Plot, in blue the points that passed all outlier detectors so far

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

# Overlay, in red, the points that remain after suppression

# Overlay, in red, the points that remain after suppression

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

```

```

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

### Suppression and Do/Undo

### Suppression and Do/Undo

The suppression object, associated with each edge is a stateful observable.  This means that other objects can observe the suppression object.  If the suppression object changes, all of the observers are notified and can take whatever action they have registered.  In addition to being observable, the suppression object keeps a history of itself.  This supports do/undo functionality (that alerts observers).

The suppression object, associated with each edge is a stateful observable.  This means that other objects can observe the suppression object.  If the suppression object changes, all of the observers are notified and can take whatever action they have registered.  In addition to being observable, the suppression object keeps a history of itself.  This supports do/undo functionality (that alerts observers).

The cell above created that object with a custom distance function.  The cells below alter $k$, the desired number of points, and $k_{error}$, the acceptable percentage of error in $k$.  These changes are then rolled back and forth.

The cell above created that object with a custom distance function.  The cells below alter $k$, the desired number of points, and $k_{error}$, the acceptable percentage of error in $k$.  These changes are then rolled back and forth.

The plotting calls remain the same for all of these example, only the first line of each is altered.

The plotting calls remain the same for all of these example, only the first line of each is altered.

--------

--------

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

$k=10$ and $k_{error}$ defaults to 10%

$k=10$ and $k_{error}$ defaults to 10%

*Take note of the bad point, in the left image, that has made it through the ratio, symmetry, and fundamental matrix computation tests.*

*Take note of the bad point, in the left image, that has made it through the ratio, symmetry, and fundamental matrix computation tests.*

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance, k=10)

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance, k=10)

# Plot, in blue the points that passed all outlier detectors so far

# Plot, in blue the points that passed all outlier detectors so far

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

# Overlay, in red, the points that remain after suppression

# Overlay, in red, the points that remain after suppression

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

```

```

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

$k = 50$

$k = 50$

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance, k=50)

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance, k=50)

# Plot, in blue the points that passed all outlier detectors so far

# Plot, in blue the points that passed all outlier detectors so far

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

# Overlay, in red, the points that remain after suppression

# Overlay, in red, the points that remain after suppression

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

```

```

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

$k=100$ and $k_{error} = 25%$

$k=100$ and $k_{error} = 25%$

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance, k=100, k_error=0.25)

#cg.edge[0][1].suppress(clean_keys=['fundamental'], func=distance, k=100, k_error=0.25)

# Plot, in blue the points that passed all outlier detectors so far

# Plot, in blue the points that passed all outlier detectors so far

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

# Overlay, in red, the points that remain after suppression

# Overlay, in red, the points that remain after suppression

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

```

```

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

Using the suppression object we can access some attributes to see how many valid points.

Using the suppression object we can access some attributes to see how many valid points.

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#cg.edge[0][1].suppression.nvalid

#cg.edge[0][1].suppression.nvalid

```

```

%% Cell type:markdown id: tags:

%% Cell type:markdown id: tags:

## Rollback

## Rollback

Now we will undo that last change.  Perhaps the error was just too high (it was not near 25% in this case, but imagine it was).

Now we will undo that last change.  Perhaps the error was just too high (it was not near 25% in this case, but imagine it was).

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

#cg.edge[0][1].suppression.rollback()

#cg.edge[0][1].suppression.rollback()

# Plot, in blue the points that passed all outlier detectors so far

# Plot, in blue the points that passed all outlier detectors so far

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

#cg.edge[0][1].plot(clean_keys=['fundamental'], line_kwargs={'linewidth':0})

# Overlay, in red, the points that remain after suppression

# Overlay, in red, the points that remain after suppression

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

#cg.edge[0][1].plot(clean_keys=['suppression'], line_kwargs={'linewidth':0}, scatter_kwargs={'color':'red'})

```

```

%% Cell type:code id: tags:

%% Cell type:code id: tags:

``` python

``` python

```

```

    "from autocnet.examples import get_path\n",

    "from autocnet.examples import get_path\n",

    "from autocnet.graph.network import CandidateGraph\n",

    "from autocnet.graph.network import CandidateGraph\n",

    "from autocnet.graph.edge import Edge\n",

    "from autocnet.graph.edge import Edge\n",

    "from autocnet.matcher.matcher import FlannMatcher\n",

    "from autocnet.matcher.feature import FlannMatcher\n",

    "\n",

    "\n",

    "from IPython.display import display\n",

    "from IPython.display import display\n",

    "\n",

    "\n",