-
Notifications
You must be signed in to change notification settings - Fork 1
/
evaluate.sh
executable file
·190 lines (143 loc) · 9.05 KB
/
evaluate.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
#!/usr/bin/tcsh -f
#####################################################################################################################################
# This code evaluates the SKL-CRM using mean per word recall, precision and number of words greater than zero (N+)
######################################################################################################################################
set YARI=$1
set DATASET_DIR=$2
set TMP_DIR=$3 # Stores temporary intermediate matrices
set DATASET_PREFIX=$4 # Test or Validation dataset for evaluation?
set BANDWIDTH=`cat $TMP_DIR/BW.txt`
set TOP_WORDS=5
set NUM_TRAIN_FEAT=`$YARI size $TMP_DIR/TRAIN_BLOBS_M | cut -d 'x' -f 1 | sed 's/ //' | cut -d ':' -f 2 | sed 's/ //'` # Number of images in our validation dataset
set NUM_TEST_IMAGES=`$YARI size $TMP_DIR/TEST_DOC_WORDS_M | cut -d 'x' -f 1 | sed 's/ //' | cut -d ':' -f 2 | sed 's/ //'` # Number of images in our validing dataset
set NUM_TRAIN_IMAGES=`$YARI size $TMP_DIR/TRAIN_DOC_WORDS_M | cut -d 'x' -f 1 | sed 's/ //' | cut -d ':' -f 2 | sed 's/ //'` # Number of images in our validing dataset
set NUM_WORDS=`$YARI size $TMP_DIR/TRAIN_DOC_WORDS_M | cut -d 'x' -f 2 | sed 's/ //'` # Number of words
set TRAIN_DOC_WORDS_FP=$DATASET_DIR/train_doc_words_DenseSift.rcv
set TEST_DOC_WORDS_FP=$DATASET_DIR/${DATASET_PREFIX}_doc_words_DenseSift.rcv
#####################################################################################################################################
# Add up the distances across each of the 15 different feature types
######################################################################################################################################
rm -rf $TMP_DIR/IMAGE_DISTS_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_GIST_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_HSV_V3H1_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_HSV_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_DS_V3H1_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_DS_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_DH_V3H1_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_DH_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_HH_V3H1_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_HH_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_HS_V3H1_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_RGB_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_HS_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_LAB_V3H1_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_LAB_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '+' $TMP_DIR/IMAGE_DISTS_RGB_V3H1_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '/' '15'
#####################################################################################################################################
# Error checking
######################################################################################################################################
$YARI print:rcv $TMP_DIR/IMAGE_DISTS_M > "$TMP_DIR/nan_find.txt"
set IS_NAN = `cat "$TMP_DIR/nan_find.txt" | grep "nan" | wc -l`
set IS_INF = `cat "$TMP_DIR/nan_find.txt" | grep "inf" | wc -l`
if ($IS_NAN > 0) then
echo "ERROR: NaN found!"
exit
endif
if ($IS_INF > 0) then
echo "ERROR: Inf found!"
exit
endif
$YARI print:rcv $TMP_DIR/WORD_PROBS_M > "$TMP_DIR/nan_find.txt"
set IS_NAN = `cat "$TMP_DIR/nan_find.txt" | grep "nan" | wc -l`
set IS_INF = `cat "$TMP_DIR/nan_find.txt" | grep "inf" | wc -l`
if ($IS_NAN > 0) then
echo "ERROR: NaN found!"
exit
endif
if ($IS_INF > 0) then
echo "ERROR: Inf found!"
exit
endif
#####################################################################################################################################
# Annotate the images
######################################################################################################################################
# Compute the image-image probabilities (P(I|J))
$YARI transpose $TMP_DIR/TRAIN_DOC_WORDS_M
$YARI $TMP_DIR/IMAGE_DISTS_M = $TMP_DIR/IMAGE_DISTS_M '/' $BANDWIDTH
$YARI $TMP_DIR/IMAGE_LOG_PROBS_M = $TMP_DIR/IMAGE_DISTS_M '#' $TMP_DIR/ID_TRAIN_FEAT_M 'logSexp'
$YARI $TMP_DIR/IMAGE_POST_PROBS_M = $TMP_DIR/ID_TEST_FEAT_M.T 'x' $TMP_DIR/IMAGE_LOG_PROBS_M
$YARI $TMP_DIR/IMAGE_PROBS_M = weigh:log2p $TMP_DIR/IMAGE_POST_PROBS_M
# Annotate the data
$YARI $TMP_DIR/ANNOTATIONS_M = $TMP_DIR/IMAGE_PROBS_M 'x' $TMP_DIR/WORD_PROBS_M # Perform the actual image annotation step
$YARI transpose $TMP_DIR/ANNOTATIONS_M
$YARI $TMP_DIR/ANNO_MAX_M = max $TMP_DIR/ANNOTATIONS_M.T
$YARI $TMP_DIR/ANNO_MIN_M = min $TMP_DIR/ANNOTATIONS_M.T
$YARI $TMP_DIR/ONES_M = ones $NUM_TEST_IMAGES 1
$YARI transpose $TMP_DIR/ANNO_MAX_M
$YARI transpose $TMP_DIR/ANNO_MIN_M
# Normalise the annotation probabilities (P(w|I)) with min-max normalisation
$YARI $TMP_DIR/ANNO_MAX_M = $TMP_DIR/ONES_M 'x' $TMP_DIR/ANNO_MAX_M.T
$YARI $TMP_DIR/ANNO_MIN_M = $TMP_DIR/ONES_M 'x' $TMP_DIR/ANNO_MIN_M.T
$YARI $TMP_DIR/ANNO_MIN_MAX_M = $TMP_DIR/ANNO_MAX_M '-' $TMP_DIR/ANNO_MIN_M
$YARI $TMP_DIR/ANNOTATIONS_M = $TMP_DIR/ANNOTATIONS_M '-' $TMP_DIR/ANNO_MIN_M
$YARI $TMP_DIR/ANNOTATIONS_M = $TMP_DIR/ANNOTATIONS_M '/' $TMP_DIR/ANNO_MIN_MAX_M
$YARI print:rcv,top=$TOP_WORDS $TMP_DIR/ANNOTATIONS_M > $TMP_DIR/ann.rcv # Print out the top 5 words for each image
#####################################################################################################################################
# Evaluate the annotations
######################################################################################################################################
echo "###################### Computing Results #######################"
###### Filter for words that occur both in training and test sets
$YARI $TMP_DIR/TEST_WORDS_FILTER_M = $TMP_DIR/TEST_DOC_WORDS_M = 1
$YARI transpose $TMP_DIR/TEST_WORDS_FILTER_M
$YARI $TMP_DIR/TEST_WORDS_FILTER_M = sum $TMP_DIR/TEST_WORDS_FILTER_M.T
$YARI $TMP_DIR/TRAIN_WORDS_FILTER_M = $TMP_DIR/TRAIN_DOC_WORDS_M = 1
$YARI transpose $TMP_DIR/TRAIN_WORDS_FILTER_M
$YARI $TMP_DIR/TRAIN_WORDS_FILTER_M = sum $TMP_DIR/TRAIN_WORDS_FILTER_M.T
$YARI $TMP_DIR/WORDS_FILTER_M = $TMP_DIR/TEST_WORDS_FILTER_M '*' $TMP_DIR/TRAIN_WORDS_FILTER_M
$YARI $TMP_DIR/WORDS_FILTER_M = $TMP_DIR/WORDS_FILTER_M = 1
$YARI transpose $TMP_DIR/WORDS_FILTER_M
$YARI $TMP_DIR/WORD_COUNT_M = sum $TMP_DIR/WORDS_FILTER_M.T
$YARI print:rcv $TMP_DIR/WORD_COUNT_M > "$TMP_DIR/word_count.txt"
set word_count = `cat "$TMP_DIR/word_count.txt" | awk '{ print $3 }'`
###### MPP #####
$YARI $TMP_DIR/TOP_ANN_M = top=5 $TMP_DIR/ANNOTATIONS_M
$YARI $TMP_DIR/RET_REL_M = $TMP_DIR/TOP_ANN_M '*' $TMP_DIR/TEST_DOC_WORDS_M
$YARI $TMP_DIR/TOP_ANN_M = $TMP_DIR/TOP_ANN_M = 1
$YARI transpose $TMP_DIR/TOP_ANN_M
$YARI $TMP_DIR/TOP_ANN_SUM_M = sum $TMP_DIR/TOP_ANN_M.T
$YARI $TMP_DIR/TOP_ANN_SUM_M = $TMP_DIR/TOP_ANN_SUM_M '*' $TMP_DIR/WORDS_FILTER_M
$YARI $TMP_DIR/RET_REL_M = $TMP_DIR/RET_REL_M = 1
$YARI transpose $TMP_DIR/RET_REL_M
$YARI $TMP_DIR/RET_REL_SUM_M = sum $TMP_DIR/RET_REL_M.T
$YARI $TMP_DIR/RET_REL_SUM_M = $TMP_DIR/RET_REL_SUM_M '*' $TMP_DIR/WORDS_FILTER_M
$YARI $TMP_DIR/PREC_M = $TMP_DIR/RET_REL_SUM_M '/' $TMP_DIR/TOP_ANN_SUM_M
$YARI transpose $TMP_DIR/PREC_M
$YARI $TMP_DIR/PREC_SUM_M = sum $TMP_DIR/PREC_M.T
$YARI $TMP_DIR/PREC_SUM_M = $TMP_DIR/PREC_SUM_M '/' $word_count
$YARI print:rcv $TMP_DIR/PREC_SUM_M > "$TMP_DIR/mpp.txt"
set mpp = `cat "$TMP_DIR/mpp.txt" | awk '{ print $3 }'`
###### MPR #####
$YARI $TMP_DIR/REL_M = $TMP_DIR/TEST_DOC_WORDS_M = 1
$YARI transpose $TMP_DIR/REL_M
$YARI $TMP_DIR/REL_SUM_M = sum $TMP_DIR/REL_M.T
$YARI $TMP_DIR/RECALL_M = $TMP_DIR/RET_REL_SUM_M '/' $TMP_DIR/REL_SUM_M
$YARI transpose $TMP_DIR/RECALL_M
$YARI $TMP_DIR/RECALL_SUM_M = sum $TMP_DIR/RECALL_M.T
$YARI $TMP_DIR/RECALL_SUM_M = $TMP_DIR/RECALL_SUM_M '/' $word_count
$YARI print:rcv $TMP_DIR/RECALL_SUM_M > "$TMP_DIR/mpr.txt"
set mpr = `cat "$TMP_DIR/mpr.txt" | awk '{ print $3 }'`
####### Words with recall > 0 ####
$YARI $TMP_DIR/RET_REL_SUM_M = $TMP_DIR/RET_REL_SUM_M = 1
$YARI transpose $TMP_DIR/RET_REL_SUM_M
$YARI $TMP_DIR/NWRGZ_M = sum $TMP_DIR/RET_REL_SUM_M.T
$YARI print:rcv $TMP_DIR/NWRGZ_M > "$TMP_DIR/nwrgz.txt"
set nrgz = `cat "$TMP_DIR/nwrgz.txt" | awk '{ print $3 }'`
set f1=`echo "2*($mpp*$mpr)/($mpp + $mpr)" | bc -l | awk '{printf "%0.4f\n", $1}'`
set res_fp=$TMP_DIR/"res.txt"
echo "Results computed on $word_count words:"
echo "MPR: " $mpr
echo "MPP: " $mpp
echo "F1: " $f1
echo "# Words recall > 0: " $nrgz
echo $f1 > $res_fp