face-landmark-detection

May 6, 2023

Importing the data from Kaggle

[1]: ! pip install -q kaggle
from google.colab import files
files.upload()
! mkdir ~/.kaggle
! cp kaggle.json ~/.kaggle/
! chmod 600 ~/.kaggle/kaggle.json

<IPython.core.display.HTML object>
Saving kaggle.json to kaggle.json

[2]: !kaggle datasets download -d drgilermo/face-images-with-marked-landmark-points

Downloading face-images-with-marked-landmark-points.zip to /content

100% 105M/105M [00:05<00:00, 23.8MB/s]
100% 105M/105M [00:05<00:00, 18.5MB/s]

[3]: !unzip face-images-with-marked-landmark-points.zip

Archive: face-images-with-marked-landmark-points.zip
inflating: face_images.npz
inflating: facial_keypoints.csv
Importing Dependencies
[4]: import numpy as np
import pandas as pd
import keras
import cv2 as cv
import xgboost as xgb
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestRegressor
from keras.layers import Conv2D, MaxPool2D, GlobalAveragePooling2D, Dense,␣
↪Dropout, Input

from keras.models import Sequential, Model

from sklearn.metrics import mean_absolute_error

1
 Loading the images
[5]: images = np.load('/content/face_images.npz')
faces = images.get(images.files[0])
faces=np.moveaxis(faces,-1,0)
faces=faces.reshape(faces.shape[0],faces.shape[1],faces.shape[1],1)

[6]: print(f"shape of the faces {faces.shape}")

shape of the faces (7049, 96, 96, 1)

[7]: df = pd.read_csv('/content/facial_keypoints.csv')

[8]: df.head()

[8]: left_eye_center_x left_eye_center_y right_eye_center_x \

0 66.033564 39.002274 30.227008
1 64.332936 34.970077 29.949277
2 65.057053 34.909642 30.903789
3 65.225739 37.261774 32.023096
4 66.725301 39.621261 32.244810

right_eye_center_y left_eye_inner_corner_x left_eye_inner_corner_y \

0 36.421678 59.582075 39.647423
1 33.448715 58.856170 35.274349
2 34.909642 59.412000 36.320968
3 37.261774 60.003339 39.127179
4 38.042032 58.565890 39.621261

left_eye_outer_corner_x left_eye_outer_corner_y right_eye_inner_corner_x \

0 73.130346 39.969997 36.356571
1 70.722723 36.187166 36.034723
2 70.984421 36.320968 37.678105
3 72.314713 38.380967 37.618643
4 72.515926 39.884466 36.982380

right_eye_inner_corner_y … nose_tip_x nose_tip_y mouth_left_corner_x \

0 37.389402 … 44.420571 57.066803 61.195308
1 34.361532 … 48.206298 55.660936 56.421447
2 36.320968 … 47.557263 53.538947 60.822947
3 38.754115 … 51.885078 54.166539 65.598887
4 39.094852 … 43.299534 64.889521 60.671411

mouth_left_corner_y mouth_right_corner_x mouth_right_corner_y \

0 79.970165 28.614496 77.388992
1 76.352000 35.122383 76.047660
2 73.014316 33.726316 72.732000
3 72.703722 37.245496 74.195478

2
 4 77.523239 31.191755 76.997301

mouth_center_top_lip_x mouth_center_top_lip_y mouth_center_bottom_lip_x \

0 43.312602 72.935459 43.130707
1 46.684596 70.266553 45.467915
2 47.274947 70.191789 47.274947
3 50.303165 70.091687 51.561183
4 44.962748 73.707387 44.227141

mouth_center_bottom_lip_y
0 84.485774
1 85.480170
2 78.659368
3 78.268383
4 86.871166

[5 rows x 30 columns]

[9]: df.corr()

[9]: left_eye_center_x left_eye_center_y \

left_eye_center_x 1.000000 0.029908
left_eye_center_y 0.029908 1.000000
right_eye_center_x 0.274459 0.259012
right_eye_center_y -0.154728 0.346331
left_eye_inner_corner_x 0.856572 0.065749
left_eye_inner_corner_y 0.022211 0.926728
left_eye_outer_corner_x 0.879976 -0.013411
left_eye_outer_corner_y 0.062128 0.956066
right_eye_inner_corner_x 0.308649 0.101996
right_eye_inner_corner_y -0.104698 0.537101
right_eye_outer_corner_x -0.065016 0.152378
right_eye_outer_corner_y -0.075857 0.229097
left_eyebrow_inner_end_x 0.631805 0.082724
left_eyebrow_inner_end_y -0.128193 0.609392
left_eyebrow_outer_end_x 0.796638 0.057346
left_eyebrow_outer_end_y -0.071141 0.760461
right_eyebrow_inner_end_x 0.268463 0.225890
right_eyebrow_inner_end_y -0.201853 0.342458
right_eyebrow_outer_end_x -0.137146 0.194586
right_eyebrow_outer_end_y -0.190548 0.070824
nose_tip_x 0.457468 -0.166891
nose_tip_y 0.192274 0.333053
mouth_left_corner_x 0.422891 -0.340755
mouth_left_corner_y 0.284087 0.480924
mouth_right_corner_x -0.008382 -0.392434
mouth_right_corner_y 0.202842 0.219366

3
mouth_center_top_lip_x 0.260753 -0.425576
mouth_center_top_lip_y 0.151302 0.251472
mouth_center_bottom_lip_x 0.374314 -0.379980
mouth_center_bottom_lip_y 0.265655 0.130062

right_eye_center_x right_eye_center_y \
left_eye_center_x 0.274459 -0.154728
left_eye_center_y 0.259012 0.346331
right_eye_center_x 1.000000 0.067621
right_eye_center_y 0.067621 1.000000
left_eye_inner_corner_x 0.294550 -0.149295
left_eye_inner_corner_y 0.127367 0.531960
left_eye_outer_corner_x -0.024408 -0.144389
left_eye_outer_corner_y 0.060942 0.221829
right_eye_inner_corner_x 0.817895 0.055758
right_eye_inner_corner_y 0.090214 0.923050
right_eye_outer_corner_x 0.861858 0.078540
right_eye_outer_corner_y 0.062292 0.954285
left_eyebrow_inner_end_x 0.301658 -0.262560
left_eyebrow_inner_end_y 0.138773 0.366292
left_eyebrow_outer_end_x -0.063364 -0.240284
left_eyebrow_outer_end_y 0.083405 0.093072
right_eyebrow_inner_end_x 0.572767 -0.121474
right_eyebrow_inner_end_y 0.137839 0.600844
right_eyebrow_outer_end_x 0.776246 -0.015320
right_eyebrow_outer_end_y 0.084007 0.751725
nose_tip_x 0.402189 0.182866
nose_tip_y 0.024502 0.329233
mouth_left_corner_x 0.061124 0.312434
mouth_left_corner_y -0.172144 0.098019
mouth_right_corner_x 0.424439 0.257956
mouth_right_corner_y -0.215123 0.397645
mouth_center_top_lip_x 0.308889 0.294145
mouth_center_top_lip_y -0.109026 0.198499
mouth_center_bottom_lip_x 0.319541 0.344010
mouth_center_bottom_lip_y -0.251105 0.044597

left_eye_inner_corner_x left_eye_inner_corner_y \
left_eye_center_x 0.856572 0.022211
left_eye_center_y 0.065749 0.926728
right_eye_center_x 0.294550 0.127367
right_eye_center_y -0.149295 0.531960
left_eye_inner_corner_x 1.000000 -0.021444
left_eye_inner_corner_y -0.021444 1.000000
left_eye_outer_corner_x 0.653318 0.010531
left_eye_outer_corner_y 0.080430 0.869743
right_eye_inner_corner_x 0.267661 0.150281

4
right_eye_inner_corner_y -0.172852 0.710354
right_eye_outer_corner_x 0.248756 0.104186
right_eye_outer_corner_y -0.106544 0.371340
left_eyebrow_inner_end_x 0.750635 -0.011428
left_eyebrow_inner_end_y -0.152820 0.677130
left_eyebrow_outer_end_x 0.622702 0.035147
left_eyebrow_outer_end_y -0.021592 0.681814
right_eyebrow_inner_end_x 0.372169 0.180705
right_eyebrow_inner_end_y -0.258597 0.489169
right_eyebrow_outer_end_x 0.112094 0.130958
right_eyebrow_outer_end_y -0.204863 0.208522
nose_tip_x 0.387317 -0.091006
nose_tip_y -0.054712 0.426897
mouth_left_corner_x 0.287427 -0.211855
mouth_left_corner_y 0.156495 0.493657
mouth_right_corner_x 0.145968 -0.335989
mouth_right_corner_y 0.046439 0.327081
mouth_center_top_lip_x 0.322217 -0.315929
mouth_center_top_lip_y -0.083635 0.387311
mouth_center_bottom_lip_x 0.242669 -0.367069
mouth_center_bottom_lip_y 0.303769 0.308555

left_eye_outer_corner_x left_eye_outer_corner_y \
left_eye_center_x 0.879976 0.062128
left_eye_center_y -0.013411 0.956066
right_eye_center_x -0.024408 0.060942
right_eye_center_y -0.144389 0.221829
left_eye_inner_corner_x 0.653318 0.080430
left_eye_inner_corner_y 0.010531 0.869743
left_eye_outer_corner_x 1.000000 0.054156
left_eye_outer_corner_y 0.054156 1.000000
right_eye_inner_corner_x 0.316254 0.050529
right_eye_inner_corner_y -0.094245 0.373183
right_eye_outer_corner_x -0.350353 0.090892
right_eye_outer_corner_y -0.077623 0.060410
left_eyebrow_inner_end_x 0.468184 0.109101
left_eyebrow_inner_end_y -0.140902 0.522014
left_eyebrow_outer_end_x 0.868225 0.139029
left_eyebrow_outer_end_y -0.097167 0.802413
right_eyebrow_inner_end_x 0.229020 0.232601
right_eyebrow_inner_end_y -0.200114 0.188524
right_eyebrow_outer_end_x -0.336098 0.164726
right_eyebrow_outer_end_y -0.213001 -0.069495
nose_tip_x 0.179189 -0.176047
nose_tip_y 0.165361 0.222928
mouth_left_corner_x 0.507755 -0.392594
mouth_left_corner_y 0.325291 0.506364

5
mouth_right_corner_x -0.136411 -0.462165
mouth_right_corner_y 0.263466 0.183239
mouth_center_top_lip_x 0.229154 -0.494482
mouth_center_top_lip_y 0.292129 0.216060
mouth_center_bottom_lip_x 0.176392 -0.561581
mouth_center_bottom_lip_y 0.332880 0.329316

right_eye_inner_corner_x right_eye_inner_corner_y \
left_eye_center_x 0.308649 -0.104698
left_eye_center_y 0.101996 0.537101
right_eye_center_x 0.817895 0.090214
right_eye_center_y 0.055758 0.923050
left_eye_inner_corner_x 0.267661 -0.172852
left_eye_inner_corner_y 0.150281 0.710354
left_eye_outer_corner_x 0.316254 -0.094245
left_eye_outer_corner_y 0.050529 0.373183
right_eye_inner_corner_x 1.000000 0.124023
right_eye_inner_corner_y 0.124023 1.000000
right_eye_outer_corner_x 0.543188 0.054442
right_eye_outer_corner_y 0.074985 0.861327
left_eyebrow_inner_end_x 0.346247 -0.243664
left_eyebrow_inner_end_y 0.137977 0.515579
left_eyebrow_outer_end_x 0.189012 -0.183551
left_eyebrow_outer_end_y 0.034023 0.237008
right_eyebrow_inner_end_x 0.657995 -0.046338
right_eyebrow_inner_end_y 0.125718 0.672297
right_eyebrow_outer_end_x 0.521745 -0.003771
right_eyebrow_outer_end_y 0.068031 0.673633
nose_tip_x 0.377931 0.053212
nose_tip_y 0.118548 0.356647
mouth_left_corner_x 0.262597 0.237858
mouth_left_corner_y -0.054888 0.225398
mouth_right_corner_x 0.262056 0.108121
mouth_right_corner_y -0.094748 0.449346
mouth_center_top_lip_x 0.369890 0.176403
mouth_center_top_lip_y 0.109785 0.354693
mouth_center_bottom_lip_x 0.309453 0.203002
mouth_center_bottom_lip_y -0.230177 0.164086

… nose_tip_x nose_tip_y mouth_left_corner_x \

left_eye_center_x … 0.457468 0.192274 0.422891
left_eye_center_y … -0.166891 0.333053 -0.340755
right_eye_center_x … 0.402189 0.024502 0.061124
right_eye_center_y … 0.182866 0.329233 0.312434
left_eye_inner_corner_x … 0.387317 -0.054712 0.287427
left_eye_inner_corner_y … -0.091006 0.426897 -0.211855
left_eye_outer_corner_x … 0.179189 0.165361 0.507755

6
left_eye_outer_corner_y … -0.176047 0.222928 -0.392594
right_eye_inner_corner_x … 0.377931 0.118548 0.262597
right_eye_inner_corner_y … 0.053212 0.356647 0.237858
right_eye_outer_corner_x … 0.156996 -0.133285 -0.177957
right_eye_outer_corner_y … 0.171852 0.118600 0.454774
left_eyebrow_inner_end_x … 0.554900 -0.087723 0.104528
left_eyebrow_inner_end_y … -0.007749 0.477947 -0.218913
left_eyebrow_outer_end_x … 0.103122 0.130824 0.320783
left_eyebrow_outer_end_y … -0.161689 0.196461 -0.438723
right_eyebrow_inner_end_x … 0.516089 0.126098 -0.059611
right_eyebrow_inner_end_y … -0.057668 0.422549 0.060252
right_eyebrow_outer_end_x … 0.046404 -0.134561 -0.302128
right_eyebrow_outer_end_y … 0.187261 0.112216 0.374869
nose_tip_x … 1.000000 0.070132 0.339788
nose_tip_y … 0.070132 1.000000 0.035640
mouth_left_corner_x … 0.339788 0.035640 1.000000
mouth_left_corner_y … -0.099397 0.442267 -0.153760
mouth_right_corner_x … 0.308948 -0.177473 0.257016
mouth_right_corner_y … 0.040560 0.415741 0.156002
mouth_center_top_lip_x … 0.703506 -0.060059 0.722451
mouth_center_top_lip_y … -0.010470 0.733981 0.119111
mouth_center_bottom_lip_x … 0.845545 0.064277 0.737955
mouth_center_bottom_lip_y … -0.010041 0.138304 0.051944

mouth_left_corner_y mouth_right_corner_x \
left_eye_center_x 0.284087 -0.008382
left_eye_center_y 0.480924 -0.392434
right_eye_center_x -0.172144 0.424439
right_eye_center_y 0.098019 0.257956
left_eye_inner_corner_x 0.156495 0.145968
left_eye_inner_corner_y 0.493657 -0.335989
left_eye_outer_corner_x 0.325291 -0.136411
left_eye_outer_corner_y 0.506364 -0.462165
right_eye_inner_corner_x -0.054888 0.262056
right_eye_inner_corner_y 0.225398 0.108121
right_eye_outer_corner_x -0.262503 0.468401
right_eye_outer_corner_y 0.031565 0.364246
left_eyebrow_inner_end_x 0.037376 0.017408
left_eyebrow_inner_end_y 0.276619 -0.170956
left_eyebrow_outer_end_x 0.389780 -0.240134
left_eyebrow_outer_end_y 0.386743 -0.385602
right_eyebrow_inner_end_x 0.139671 0.024607
right_eyebrow_inner_end_y 0.099020 0.117008
right_eyebrow_outer_end_x -0.245991 0.315302
right_eyebrow_outer_end_y -0.066545 0.351068
nose_tip_x -0.099397 0.308948
nose_tip_y 0.442267 -0.177473

7
mouth_left_corner_x -0.153760 0.257016
mouth_left_corner_y 1.000000 -0.231965
mouth_right_corner_x -0.231965 1.000000
mouth_right_corner_y 0.845165 0.047169
mouth_center_top_lip_x -0.237070 0.706419
mouth_center_top_lip_y 0.726913 -0.182543
mouth_center_bottom_lip_x -0.278776 0.727747
mouth_center_bottom_lip_y 0.724428 -0.183575

mouth_right_corner_y mouth_center_top_lip_x \
left_eye_center_x 0.202842 0.260753
left_eye_center_y 0.219366 -0.425576
right_eye_center_x -0.215123 0.308889
right_eye_center_y 0.397645 0.294145
left_eye_inner_corner_x 0.046439 0.322217
left_eye_inner_corner_y 0.327081 -0.315929
left_eye_outer_corner_x 0.263466 0.229154
left_eye_outer_corner_y 0.183239 -0.494482
right_eye_inner_corner_x -0.094748 0.369890
right_eye_inner_corner_y 0.449346 0.176403
right_eye_outer_corner_x -0.323196 0.179465
right_eye_outer_corner_y 0.398740 0.451419
left_eyebrow_inner_end_x -0.124903 0.266158
left_eyebrow_inner_end_y 0.202403 -0.183568
left_eyebrow_outer_end_x 0.264717 0.058748
left_eyebrow_outer_end_y 0.100390 -0.448908
right_eyebrow_inner_end_x -0.035666 0.191600
right_eyebrow_inner_end_y 0.244345 0.058862
right_eyebrow_outer_end_x -0.383826 -0.001356
right_eyebrow_outer_end_y 0.276684 0.426359
nose_tip_x 0.040560 0.703506
nose_tip_y 0.415741 -0.060059
mouth_left_corner_x 0.156002 0.722451
mouth_left_corner_y 0.845165 -0.237070
mouth_right_corner_x 0.047169 0.706419
mouth_right_corner_y 1.000000 0.114100
mouth_center_top_lip_x 0.114100 1.000000
mouth_center_top_lip_y 0.741801 -0.021879
mouth_center_bottom_lip_x 0.119912 0.975837
mouth_center_bottom_lip_y 0.681152 -0.077797

mouth_center_top_lip_y mouth_center_bottom_lip_x \
left_eye_center_x 0.151302 0.374314
left_eye_center_y 0.251472 -0.379980
right_eye_center_x -0.109026 0.319541
right_eye_center_y 0.198499 0.344010
left_eye_inner_corner_x -0.083635 0.242669

8
left_eye_inner_corner_y 0.387311 -0.367069
left_eye_outer_corner_x 0.292129 0.176392
left_eye_outer_corner_y 0.216060 -0.561581
right_eye_inner_corner_x 0.109785 0.309453
right_eye_inner_corner_y 0.354693 0.203002
right_eye_outer_corner_x -0.327911 0.140704
right_eye_outer_corner_y 0.139957 0.504343
left_eyebrow_inner_end_x -0.114208 0.178256
left_eyebrow_inner_end_y 0.414661 -0.219322
left_eyebrow_outer_end_x 0.243773 0.020986
left_eyebrow_outer_end_y 0.159086 -0.493071
right_eyebrow_inner_end_x 0.081940 0.093641
right_eyebrow_inner_end_y 0.371001 0.076888
right_eyebrow_outer_end_x -0.271817 -0.042747
right_eyebrow_outer_end_y 0.085133 0.477362
nose_tip_x -0.010470 0.845545
nose_tip_y 0.733981 0.064277
mouth_left_corner_x 0.119111 0.737955
mouth_left_corner_y 0.726913 -0.278776
mouth_right_corner_x -0.182543 0.727747
mouth_right_corner_y 0.741801 0.119912
mouth_center_top_lip_x -0.021879 0.975837
mouth_center_top_lip_y 1.000000 -0.047274
mouth_center_bottom_lip_x -0.047274 1.000000
mouth_center_bottom_lip_y 0.410089 -0.034845

mouth_center_bottom_lip_y
left_eye_center_x 0.265655
left_eye_center_y 0.130062
right_eye_center_x -0.251105
right_eye_center_y 0.044597
left_eye_inner_corner_x 0.303769
left_eye_inner_corner_y 0.308555
left_eye_outer_corner_x 0.332880
left_eye_outer_corner_y 0.329316
right_eye_inner_corner_x -0.230177
right_eye_inner_corner_y 0.164086
right_eye_outer_corner_x -0.277736
right_eye_outer_corner_y 0.090795
left_eyebrow_inner_end_x 0.079219
left_eyebrow_inner_end_y 0.103497
left_eyebrow_outer_end_x 0.419651
left_eyebrow_outer_end_y 0.234645
right_eyebrow_inner_end_x -0.016916
right_eyebrow_inner_end_y -0.019619
right_eyebrow_outer_end_x -0.368392
right_eyebrow_outer_end_y 0.018289

9
 nose_tip_x -0.010041
nose_tip_y 0.138304
mouth_left_corner_x 0.051944
mouth_left_corner_y 0.724428
mouth_right_corner_x -0.183575
mouth_right_corner_y 0.681152
mouth_center_top_lip_x -0.077797
mouth_center_top_lip_y 0.410089
mouth_center_bottom_lip_x -0.034845
mouth_center_bottom_lip_y 1.000000

[30 rows x 30 columns]

[10]: df.isnull().sum()

[10]: left_eye_center_x 10
left_eye_center_y 10
right_eye_center_x 13
right_eye_center_y 13
left_eye_inner_corner_x 4778
left_eye_inner_corner_y 4778
left_eye_outer_corner_x 4782
left_eye_outer_corner_y 4782
right_eye_inner_corner_x 4781
right_eye_inner_corner_y 4781
right_eye_outer_corner_x 4781
right_eye_outer_corner_y 4781
left_eyebrow_inner_end_x 4779
left_eyebrow_inner_end_y 4779
left_eyebrow_outer_end_x 4824
left_eyebrow_outer_end_y 4824
right_eyebrow_inner_end_x 4779
right_eyebrow_inner_end_y 4779
right_eyebrow_outer_end_x 4813
right_eyebrow_outer_end_y 4813
nose_tip_x 0
nose_tip_y 0
mouth_left_corner_x 4780
mouth_left_corner_y 4780
mouth_right_corner_x 4779
mouth_right_corner_y 4779
mouth_center_top_lip_x 4774
mouth_center_top_lip_y 4774
mouth_center_bottom_lip_x 33
mouth_center_bottom_lip_y 33
dtype: int64

10
[11]: def missing_values(df, column): # handling the missing values with mean
df[column] = df[column].fillna(df[column].mean())

[12]: for i in df.columns:

missing_values(df, i)

[13]: df.isnull().sum()

[13]: left_eye_center_x 0
left_eye_center_y 0
right_eye_center_x 0
right_eye_center_y 0
left_eye_inner_corner_x 0
left_eye_inner_corner_y 0
left_eye_outer_corner_x 0
left_eye_outer_corner_y 0
right_eye_inner_corner_x 0
right_eye_inner_corner_y 0
right_eye_outer_corner_x 0
right_eye_outer_corner_y 0
left_eyebrow_inner_end_x 0
left_eyebrow_inner_end_y 0
left_eyebrow_outer_end_x 0
left_eyebrow_outer_end_y 0
right_eyebrow_inner_end_x 0
right_eyebrow_inner_end_y 0
right_eyebrow_outer_end_x 0
right_eyebrow_outer_end_y 0
nose_tip_x 0
nose_tip_y 0
mouth_left_corner_x 0
mouth_left_corner_y 0
mouth_right_corner_x 0
mouth_right_corner_y 0
mouth_center_top_lip_x 0
mouth_center_top_lip_y 0
mouth_center_bottom_lip_x 0
mouth_center_bottom_lip_y 0
dtype: int64

[14]: face_points = [] # appending the landmarks into face_points list

for i in range(df.shape[0]):
face_points.append(list(df.iloc[i,:]))

[15]: face_points = np.array(face_points) #converting the face points into numpy␣

↪array

11
[16]: plt.figure(figsize=(10,5))
for i in range(25):
plt.subplot(5,5,i+1)
plt.imshow(faces[i], cmap='gray')
plt.axis("off")
plt.show()

[17]: def plot_points(img_number):

points = list(df.iloc[img_number,:])
plt.imshow(faces[img_number], cmap='gray')
for i in range(0, len(points), 2):
plt.scatter(points[i:i+2][0], points[i:i+2][1], c='y')
plt.show()
plot_points(799)

12
[18]: faces = faces/255 # normalising the faces

[19]: x_train, x_test, y_train, y_test = train_test_split(faces, face_points,␣

↪test_size = 0.2, random_state=44)

[20]: print(f"the shape of x_train is {x_train.shape}")

print(f"the shape of y_train is {y_train.shape}")
print(f"the shape of x_test is {x_test.shape}")
print(f"the shape of y_test is {y_test.shape}")

the shape of x_train is (5639, 96, 96, 1)

the shape of y_train is (5639, 30)
the shape of x_test is (1410, 96, 96, 1)
the shape of y_test is (1410, 30)
CNN Model
[21]: model = Sequential()

model.add(Conv2D(32, (3,3), input_shape=(96,96,1), activation="relu",␣

↪padding="same"))

model.add(MaxPool2D(2,2))
model.add(Dropout(0.1))

13
 model.add(Conv2D(64, (3,3), activation="relu",padding="same"))
model.add(MaxPool2D(2,2))
model.add(Dropout(0.2))
model.add(Conv2D(128, (3,3), activation="relu", padding="same"))
model.add(MaxPool2D(2,2))
model.add(Dropout(0.3))
model.add(Conv2D(256, (3,3), activation="relu", padding="same"))
model.add(MaxPool2D(2,2))
model.add(Dropout(0.3))
model.add(GlobalAveragePooling2D())
model.add(Dense(128, activation="relu"))
model.add(Dense(30, activation="leaky_relu"))

[22]: model.summary()

Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d (Conv2D) (None, 96, 96, 32) 320

max_pooling2d (MaxPooling2D (None, 48, 48, 32) 0

)

dropout (Dropout) (None, 48, 48, 32) 0

conv2d_1 (Conv2D) (None, 48, 48, 64) 18496

max_pooling2d_1 (MaxPooling (None, 24, 24, 64) 0

2D)

dropout_1 (Dropout) (None, 24, 24, 64) 0

conv2d_2 (Conv2D) (None, 24, 24, 128) 73856

max_pooling2d_2 (MaxPooling (None, 12, 12, 128) 0

2D)

dropout_2 (Dropout) (None, 12, 12, 128) 0

conv2d_3 (Conv2D) (None, 12, 12, 256) 295168

max_pooling2d_3 (MaxPooling (None, 6, 6, 256) 0

2D)

dropout_3 (Dropout) (None, 6, 6, 256) 0

14
 global_average_pooling2d (G (None, 256) 0
lobalAveragePooling2D)

dense (Dense) (None, 128) 32896

dense_1 (Dense) (None, 30) 3870

=================================================================
Total params: 424,606
Trainable params: 424,606
Non-trainable params: 0
_________________________________________________________________

[23]: model.compile(optimizer="adam", loss="mse", metrics=['mae'])

[24]: history = model.fit(x_train, y_train, validation_data=(x_test, y_test),␣

↪epochs=100)

Epoch 1/100
177/177 [==============================] - 15s 25ms/step - loss: 247.6686 - mae:
8.0113 - val_loss: 83.7798 - val_mae: 8.1261
Epoch 2/100
177/177 [==============================] - 3s 17ms/step - loss: 10.6864 - mae:
2.1842 - val_loss: 113.7821 - val_mae: 9.6577
Epoch 3/100
177/177 [==============================] - 3s 16ms/step - loss: 9.2263 - mae:
1.9392 - val_loss: 110.1818 - val_mae: 9.4742
Epoch 4/100
177/177 [==============================] - 3s 16ms/step - loss: 9.3054 - mae:
1.9498 - val_loss: 98.8887 - val_mae: 8.9893
Epoch 5/100
177/177 [==============================] - 3s 16ms/step - loss: 9.6848 - mae:
2.0189 - val_loss: 103.1080 - val_mae: 9.1244
Epoch 6/100
177/177 [==============================] - 3s 16ms/step - loss: 8.8998 - mae:
1.8674 - val_loss: 159.7516 - val_mae: 11.5458
Epoch 7/100
177/177 [==============================] - 3s 17ms/step - loss: 9.4427 - mae:
1.9738 - val_loss: 111.6284 - val_mae: 9.5531
Epoch 8/100
177/177 [==============================] - 3s 17ms/step - loss: 9.4034 - mae:
1.9845 - val_loss: 97.8728 - val_mae: 8.9150
Epoch 9/100
177/177 [==============================] - 3s 16ms/step - loss: 9.5542 - mae:
1.9968 - val_loss: 133.2152 - val_mae: 10.4973
Epoch 10/100
177/177 [==============================] - 3s 16ms/step - loss: 9.3016 - mae:
1.9448 - val_loss: 100.8673 - val_mae: 9.1203

15
Epoch 11/100
177/177 [==============================] - 3s 17ms/step - loss: 9.5358 - mae:
1.9911 - val_loss: 89.7404 - val_mae: 8.5451
Epoch 12/100
177/177 [==============================] - 3s 16ms/step - loss: 9.2227 - mae:
1.9397 - val_loss: 106.7844 - val_mae: 9.3419
Epoch 13/100
177/177 [==============================] - 3s 16ms/step - loss: 8.8619 - mae:
1.8504 - val_loss: 102.1288 - val_mae: 9.1179
Epoch 14/100
177/177 [==============================] - 3s 16ms/step - loss: 9.2572 - mae:
1.9336 - val_loss: 106.6310 - val_mae: 9.3546
Epoch 15/100
177/177 [==============================] - 3s 17ms/step - loss: 9.3389 - mae:
1.9620 - val_loss: 82.9103 - val_mae: 8.1393
Epoch 16/100
177/177 [==============================] - 3s 17ms/step - loss: 9.1877 - mae:
1.9333 - val_loss: 90.3143 - val_mae: 8.5493
Epoch 17/100
177/177 [==============================] - 3s 16ms/step - loss: 9.1932 - mae:
1.9149 - val_loss: 114.8879 - val_mae: 9.7004
Epoch 18/100
177/177 [==============================] - 3s 16ms/step - loss: 9.1223 - mae:
1.9126 - val_loss: 84.8320 - val_mae: 8.2596
Epoch 19/100
177/177 [==============================] - 3s 17ms/step - loss: 9.2219 - mae:
1.9320 - val_loss: 103.1709 - val_mae: 9.1269
Epoch 20/100
177/177 [==============================] - 3s 16ms/step - loss: 8.8229 - mae:
1.8510 - val_loss: 101.4413 - val_mae: 9.1014
Epoch 21/100
177/177 [==============================] - 3s 17ms/step - loss: 8.6790 - mae:
1.8219 - val_loss: 65.5756 - val_mae: 7.2342
Epoch 22/100
177/177 [==============================] - 3s 17ms/step - loss: 9.1055 - mae:
1.9128 - val_loss: 79.3852 - val_mae: 7.9963
Epoch 23/100
177/177 [==============================] - 3s 17ms/step - loss: 8.9796 - mae:
1.8918 - val_loss: 77.5703 - val_mae: 7.9403
Epoch 24/100
177/177 [==============================] - 3s 17ms/step - loss: 8.7981 - mae:
1.8426 - val_loss: 73.4912 - val_mae: 7.6473
Epoch 25/100
177/177 [==============================] - 3s 17ms/step - loss: 8.8405 - mae:
1.8562 - val_loss: 95.9177 - val_mae: 8.8139
Epoch 26/100
177/177 [==============================] - 3s 17ms/step - loss: 8.7959 - mae:
1.8459 - val_loss: 101.8479 - val_mae: 9.1248

16
Epoch 27/100
177/177 [==============================] - 3s 17ms/step - loss: 8.9323 - mae:
1.8700 - val_loss: 53.6439 - val_mae: 6.5124
Epoch 28/100
177/177 [==============================] - 3s 17ms/step - loss: 8.7616 - mae:
1.8328 - val_loss: 67.9234 - val_mae: 7.3416
Epoch 29/100
177/177 [==============================] - 3s 16ms/step - loss: 9.0021 - mae:
1.8954 - val_loss: 97.4173 - val_mae: 8.8958
Epoch 30/100
177/177 [==============================] - 3s 16ms/step - loss: 8.9739 - mae:
1.8845 - val_loss: 74.2736 - val_mae: 7.6851
Epoch 31/100
177/177 [==============================] - 3s 17ms/step - loss: 8.7079 - mae:
1.8233 - val_loss: 37.1465 - val_mae: 5.2425
Epoch 32/100
177/177 [==============================] - 3s 17ms/step - loss: 8.8279 - mae:
1.8487 - val_loss: 52.3115 - val_mae: 6.3940
Epoch 33/100
177/177 [==============================] - 3s 16ms/step - loss: 8.7103 - mae:
1.8222 - val_loss: 69.0528 - val_mae: 7.4193
Epoch 34/100
177/177 [==============================] - 3s 17ms/step - loss: 8.7592 - mae:
1.8341 - val_loss: 82.6843 - val_mae: 8.1918
Epoch 35/100
177/177 [==============================] - 3s 17ms/step - loss: 8.5312 - mae:
1.7767 - val_loss: 69.9608 - val_mae: 7.4930
Epoch 36/100
177/177 [==============================] - 3s 17ms/step - loss: 8.5891 - mae:
1.8035 - val_loss: 61.9837 - val_mae: 7.0418
Epoch 37/100
177/177 [==============================] - 3s 16ms/step - loss: 8.7157 - mae:
1.8235 - val_loss: 78.1996 - val_mae: 7.9179
Epoch 38/100
177/177 [==============================] - 3s 17ms/step - loss: 8.7549 - mae:
1.8291 - val_loss: 66.0679 - val_mae: 7.2954
Epoch 39/100
177/177 [==============================] - 3s 16ms/step - loss: 8.7454 - mae:
1.8421 - val_loss: 65.0089 - val_mae: 7.1298
Epoch 40/100
177/177 [==============================] - 3s 17ms/step - loss: 8.7682 - mae:
1.8444 - val_loss: 61.1667 - val_mae: 6.9679
Epoch 41/100
177/177 [==============================] - 3s 16ms/step - loss: 8.4494 - mae:
1.7594 - val_loss: 54.8232 - val_mae: 6.5255
Epoch 42/100
177/177 [==============================] - 3s 16ms/step - loss: 9.3137 - mae:
1.9588 - val_loss: 44.4323 - val_mae: 5.8153

17
Epoch 43/100
177/177 [==============================] - 3s 16ms/step - loss: 8.9189 - mae:
1.8762 - val_loss: 54.5603 - val_mae: 6.5240
Epoch 44/100
177/177 [==============================] - 3s 17ms/step - loss: 8.6738 - mae:
1.8216 - val_loss: 72.8856 - val_mae: 7.6386
Epoch 45/100
177/177 [==============================] - 3s 16ms/step - loss: 8.5333 - mae:
1.7914 - val_loss: 62.3959 - val_mae: 6.9697
Epoch 46/100
177/177 [==============================] - 3s 16ms/step - loss: 8.7475 - mae:
1.8180 - val_loss: 47.0339 - val_mae: 6.0760
Epoch 47/100
177/177 [==============================] - 3s 17ms/step - loss: 8.5913 - mae:
1.8032 - val_loss: 47.9425 - val_mae: 6.1126
Epoch 48/100
177/177 [==============================] - 3s 16ms/step - loss: 8.4712 - mae:
1.7779 - val_loss: 56.3800 - val_mae: 6.6307
Epoch 49/100
177/177 [==============================] - 3s 17ms/step - loss: 8.4064 - mae:
1.7641 - val_loss: 53.0067 - val_mae: 6.4272
Epoch 50/100
177/177 [==============================] - 3s 16ms/step - loss: 8.4607 - mae:
1.7763 - val_loss: 63.2502 - val_mae: 7.0850
Epoch 51/100
177/177 [==============================] - 3s 17ms/step - loss: 8.5722 - mae:
1.8009 - val_loss: 52.5184 - val_mae: 6.4305
Epoch 52/100
177/177 [==============================] - 3s 16ms/step - loss: 8.3858 - mae:
1.7621 - val_loss: 53.6307 - val_mae: 6.5158
Epoch 53/100
177/177 [==============================] - 3s 17ms/step - loss: 8.2413 - mae:
1.7310 - val_loss: 53.4798 - val_mae: 6.4642
Epoch 54/100
177/177 [==============================] - 3s 16ms/step - loss: 8.5102 - mae:
1.7841 - val_loss: 44.5038 - val_mae: 5.8885
Epoch 55/100
177/177 [==============================] - 3s 16ms/step - loss: 8.6417 - mae:
1.8303 - val_loss: 52.5037 - val_mae: 6.4289
Epoch 56/100
177/177 [==============================] - 3s 16ms/step - loss: 8.4937 - mae:
1.7833 - val_loss: 52.7772 - val_mae: 6.4411
Epoch 57/100
177/177 [==============================] - 3s 17ms/step - loss: 8.2668 - mae:
1.7316 - val_loss: 32.1862 - val_mae: 4.8185
Epoch 58/100
177/177 [==============================] - 3s 16ms/step - loss: 8.7145 - mae:
1.8418 - val_loss: 26.5276 - val_mae: 4.3258

18
Epoch 59/100
177/177 [==============================] - 3s 16ms/step - loss: 8.2396 - mae:
1.7212 - val_loss: 35.2654 - val_mae: 5.1483
Epoch 60/100
177/177 [==============================] - 3s 16ms/step - loss: 8.4176 - mae:
1.7674 - val_loss: 43.3570 - val_mae: 5.7617
Epoch 61/100
177/177 [==============================] - 3s 17ms/step - loss: 8.5178 - mae:
1.7926 - val_loss: 42.6054 - val_mae: 5.7153
Epoch 62/100
177/177 [==============================] - 3s 16ms/step - loss: 8.3288 - mae:
1.7395 - val_loss: 33.5896 - val_mae: 5.0125
Epoch 63/100
177/177 [==============================] - 3s 16ms/step - loss: 8.1065 - mae:
1.6984 - val_loss: 20.0270 - val_mae: 3.5657
Epoch 64/100
177/177 [==============================] - 3s 16ms/step - loss: 8.3581 - mae:
1.7635 - val_loss: 22.6849 - val_mae: 3.9956
Epoch 65/100
177/177 [==============================] - 3s 16ms/step - loss: 8.0364 - mae:
1.6897 - val_loss: 26.7727 - val_mae: 4.3404
Epoch 66/100
177/177 [==============================] - 3s 16ms/step - loss: 8.1213 - mae:
1.7080 - val_loss: 21.6563 - val_mae: 3.8046
Epoch 67/100
177/177 [==============================] - 3s 15ms/step - loss: 8.2241 - mae:
1.7429 - val_loss: 16.0141 - val_mae: 3.2331
Epoch 68/100
177/177 [==============================] - 3s 16ms/step - loss: 8.0199 - mae:
1.6879 - val_loss: 42.4361 - val_mae: 5.7043
Epoch 69/100
177/177 [==============================] - 3s 15ms/step - loss: 8.3339 - mae:
1.7552 - val_loss: 23.3266 - val_mae: 4.0400
Epoch 70/100
177/177 [==============================] - 3s 17ms/step - loss: 8.1687 - mae:
1.7268 - val_loss: 16.2006 - val_mae: 3.2437
Epoch 71/100
177/177 [==============================] - 3s 16ms/step - loss: 8.1271 - mae:
1.7238 - val_loss: 15.0111 - val_mae: 2.9976
Epoch 72/100
177/177 [==============================] - 3s 16ms/step - loss: 8.0443 - mae:
1.7159 - val_loss: 12.3291 - val_mae: 2.6578
Epoch 73/100
177/177 [==============================] - 3s 16ms/step - loss: 8.0654 - mae:
1.7158 - val_loss: 24.9140 - val_mae: 4.2030
Epoch 74/100
177/177 [==============================] - 3s 17ms/step - loss: 8.0045 - mae:
1.7041 - val_loss: 16.1755 - val_mae: 3.2555

19
Epoch 75/100
177/177 [==============================] - 3s 15ms/step - loss: 7.8202 - mae:
1.6691 - val_loss: 20.4917 - val_mae: 3.7420
Epoch 76/100
177/177 [==============================] - 3s 15ms/step - loss: 8.0126 - mae:
1.7227 - val_loss: 19.7065 - val_mae: 3.6758
Epoch 77/100
177/177 [==============================] - 3s 15ms/step - loss: 7.8310 - mae:
1.6813 - val_loss: 15.4774 - val_mae: 3.1098
Epoch 78/100
177/177 [==============================] - 3s 17ms/step - loss: 7.7742 - mae:
1.6670 - val_loss: 16.0787 - val_mae: 3.1778
Epoch 79/100
177/177 [==============================] - 3s 17ms/step - loss: 7.8609 - mae:
1.6908 - val_loss: 21.0233 - val_mae: 3.7739
Epoch 80/100
177/177 [==============================] - 3s 15ms/step - loss: 7.8045 - mae:
1.6851 - val_loss: 17.0281 - val_mae: 3.3428
Epoch 81/100
177/177 [==============================] - 3s 15ms/step - loss: 7.8090 - mae:
1.7006 - val_loss: 13.2803 - val_mae: 2.8194
Epoch 82/100
177/177 [==============================] - 3s 16ms/step - loss: 8.0213 - mae:
1.7467 - val_loss: 11.8131 - val_mae: 2.6356
Epoch 83/100
177/177 [==============================] - 3s 17ms/step - loss: 7.7927 - mae:
1.7003 - val_loss: 12.2382 - val_mae: 2.6739
Epoch 84/100
177/177 [==============================] - 3s 16ms/step - loss: 7.6791 - mae:
1.6584 - val_loss: 15.3685 - val_mae: 3.1307
Epoch 85/100
177/177 [==============================] - 3s 16ms/step - loss: 7.6913 - mae:
1.6834 - val_loss: 16.1119 - val_mae: 3.2295
Epoch 86/100
177/177 [==============================] - 3s 15ms/step - loss: 7.5466 - mae:
1.6398 - val_loss: 10.7598 - val_mae: 2.4090
Epoch 87/100
177/177 [==============================] - 3s 16ms/step - loss: 7.6180 - mae:
1.6670 - val_loss: 13.5462 - val_mae: 2.8233
Epoch 88/100
177/177 [==============================] - 3s 15ms/step - loss: 7.7813 - mae:
1.7066 - val_loss: 18.1851 - val_mae: 3.5273
Epoch 89/100
177/177 [==============================] - 3s 15ms/step - loss: 7.6687 - mae:
1.6926 - val_loss: 18.3210 - val_mae: 3.5165
Epoch 90/100
177/177 [==============================] - 3s 16ms/step - loss: 7.7163 - mae:
1.6930 - val_loss: 13.8138 - val_mae: 2.9126

20
 Epoch 91/100
177/177 [==============================] - 3s 17ms/step - loss: 7.3981 - mae:
1.6206 - val_loss: 8.9744 - val_mae: 2.1093
Epoch 92/100
177/177 [==============================] - 3s 17ms/step - loss: 7.5680 - mae:
1.6752 - val_loss: 19.7709 - val_mae: 3.6264
Epoch 93/100
177/177 [==============================] - 3s 15ms/step - loss: 7.6256 - mae:
1.6861 - val_loss: 13.3184 - val_mae: 2.8322
Epoch 94/100
177/177 [==============================] - 3s 16ms/step - loss: 7.3932 - mae:
1.6445 - val_loss: 10.0174 - val_mae: 2.2282
Epoch 95/100
177/177 [==============================] - 3s 17ms/step - loss: 7.3429 - mae:
1.6242 - val_loss: 11.8309 - val_mae: 2.6271
Epoch 96/100
177/177 [==============================] - 3s 16ms/step - loss: 7.5236 - mae:
1.6728 - val_loss: 12.2052 - val_mae: 2.6475
Epoch 97/100
177/177 [==============================] - 3s 16ms/step - loss: 7.4714 - mae:
1.6601 - val_loss: 7.3549 - val_mae: 1.6120
Epoch 98/100
177/177 [==============================] - 3s 16ms/step - loss: 7.4413 - mae:
1.6529 - val_loss: 8.3854 - val_mae: 1.9461
Epoch 99/100
177/177 [==============================] - 3s 16ms/step - loss: 7.4853 - mae:
1.6658 - val_loss: 11.0050 - val_mae: 2.4938
Epoch 100/100
177/177 [==============================] - 3s 17ms/step - loss: 7.2580 - mae:
1.6183 - val_loss: 11.8805 - val_mae: 2.6062

[25]: y_pred = model.predict(x_test)

45/45 [==============================] - 0s 4ms/step

[26]: print(f"mean absolute error of the model is {mean_absolute_error(y_pred,␣

↪y_test)}")

mean absolute error of the model is 2.606207433861229

[27]: def plot_points(img_number):

plt.imshow(x_test[img_number], cmap='gray')
predicted_points = model.predict(x_test[img_number].reshape(1, 96, 96, 1))
predicted_points = predicted_points[0]
original_points = y_test[img_number]
for i in range(0, 30, 2):
plt.scatter(original_points[i:i+2][0], original_points[i:i+2][1], c='y') ␣
↪ #original

21
 plt.scatter(predicted_points[i:i+2][0], predicted_points[i:i+2][1], c='r') ␣
↪ #predicted
plt.show()
plot_points(555)

1/1 [==============================] - 0s 139ms/step

CNN Model as Feature Extractor + Random Forest for Regression

[29]: features_model = Model(inputs = [model.input], outputs = [model.
↪get_layer("global_average_pooling2d").output])

#creating a model with above CNN model upto GlobalAveragePooling2D layer output

[30]: train_features = features_model.predict(x_train) # predicting the features on␣

↪x_train

177/177 [==============================] - 1s 3ms/step

[31]: rf_reg = RandomForestRegressor()

[32]: rf_reg.fit(train_features, y_train) # fitting random forest on train_features␣

↪and y_train

22
[32]: RandomForestRegressor()

[33]: test_features = features_model.predict(x_test) # predicting the test features

45/45 [==============================] - 0s 3ms/step

[34]: output_points = rf_reg.predict(test_features) # predicting the output features

[35]: print(f"the mean abosulute error for random forest regressor is␣
↪{mean_absolute_error(output_points, y_test)}")

the mean abosulute error for random forest regressor is 1.2129046985875478

[36]: def plot_output(img_number):

predicted_points = output_points[img_number]
original_points = y_test[img_number]
plt.subplot(1,2,1)
plt.imshow(x_test[img_number], cmap='gray')
for i in range(0, 30, 2):
plt.scatter(original_points[i:i+2][0], original_points[i:i+2][1], c='y')
plt.subplot(1,2,2)
plt.imshow(x_test[img_number], cmap='gray')
for i in range(0, 30, 2):
plt.scatter(predicted_points[i:i+2][0], predicted_points[i:i+2][1], c='r')
plt.show()
plot_output(555)

XGBoost as the Regressor

[37]: xgb_reg = xgb.XGBRegressor()

23
[38]: xgb_reg.fit(train_features, y_train) # training xgboost on train features

[38]: XGBRegressor(base_score=None, booster=None, callbacks=None,

colsample_bylevel=None, colsample_bynode=None,
colsample_bytree=None, early_stopping_rounds=None,
enable_categorical=False, eval_metric=None, feature_types=None,
gamma=None, gpu_id=None, grow_policy=None, importance_type=None,
interaction_constraints=None, learning_rate=None, max_bin=None,
max_cat_threshold=None, max_cat_to_onehot=None,
max_delta_step=None, max_depth=None, max_leaves=None,
min_child_weight=None, missing=nan, monotone_constraints=None,
n_estimators=100, n_jobs=None, num_parallel_tree=None,
predictor=None, random_state=None, …)

[40]: y_pred_xgb = xgb_reg.predict(test_features)

[41]: print(f"the mean abosulute error for xgboost regressor is␣

↪{mean_absolute_error(y_pred_xgb, y_test)}")

the mean abosulute error for xgboost regressor is 1.331619245343443

[42]: def plot_output(img_number):

predicted_points = y_pred_xgb[img_number]
original_points = y_test[img_number]
plt.subplot(1,2,1)
plt.imshow(x_test[img_number], cmap='gray')
for i in range(0, 30, 2):
plt.scatter(original_points[i:i+2][0], original_points[i:i+2][1], c='y') ␣
↪# original

plt.subplot(1,2,2)
plt.imshow(x_test[img_number], cmap='gray')
for i in range(0, 30, 2):
plt.scatter(predicted_points[i:i+2][0], predicted_points[i:i+2][1], c='r') ␣
↪ # predicted

plt.show()
plot_output(555)

24
 Comparing the 3 Models
[56]: def compare(img_number): # plotting the comparisons
y_pred_cnn = y_pred[img_number]
y_pred_rf = output_points[img_number]
y_pred_Xgb = y_pred_xgb[img_number]
y_original = y_test[img_number]

plt.figure(figsize=(15,10))
plt.subplot(1, 4, 1)
plt.imshow(x_test[img_number], cmap='gray')
plt.title("Original Landmarks")
plt.axis("off")
plt.imshow(x_test[img_number], cmap='gray')
for i in range(0, 30, 2):
plt.scatter(y_original[i:i+2][0], y_original[i:i+2][1], c='y')

plt.subplot(1, 4, 2)
plt.imshow(x_test[img_number], cmap='gray')
plt.title("CNN prediction")
plt.axis("off")
for i in range(0, 30, 2):
plt.scatter(y_pred_cnn[i:i+2][0], y_pred_cnn[i:i+2][1], c='y')

plt.subplot(1, 4, 3)
plt.imshow(x_test[img_number], cmap='gray')
plt.title("Random Forest prediction")
plt.axis("off")
for i in range(0, 30, 2):
plt.scatter(y_pred_rf[i:i+2][0], y_pred_rf[i:i+2][1], c='y')

25
 plt.subplot(1, 4, 4)
plt.imshow(x_test[img_number], cmap='gray')
plt.title("XGBoost prediction")
plt.axis("off")
for i in range(0, 30, 2):
plt.scatter(y_pred_Xgb[i:i+2][0], y_pred_Xgb[i:i+2][1], c='y')
plt.show()

compare(23)

[ ]:

26