update eval

.gitignore

@@ -166,4 +166,5 @@ data/
 content-based-*.ipynb
 cbir_dev.ipynb
 *.ipynb
-*.png
+*.png
+*.out

cbir/metrics.py

@@ -1,9 +1,14 @@
 def average_precision(predictions, ground_truths, k):
     top_k_predictions = predictions[:k]
 
-    relevant_items = sum(1 for pred in top_k_predictions if pred in ground_truths)
+    hit_count = 1
+    relevant_items = []
+    for i, pred in enumerate(top_k_predictions):
+        if pred in ground_truths:
+            relevant_items.append(hit_count/(i+1))
+            hit_count += 1
 
-    ap = relevant_items / len(top_k_predictions)
+    ap = sum(relevant_items) / hit_count
 
     return ap
 
@@ -12,7 +17,8 @@ def hit_rate(predictions, ground_truths, k):
 
     relevant_items = sum(1 for pred in top_k_predictions if pred in ground_truths)
 
-    hit = 1 if relevant_items >= 1 else 0
+    # hit = 1 if relevant_items >= 1 else 0
+    hit = relevant_items / len(top_k_predictions)
 
     return hit
 

ensemble_all.py

@@ -61,7 +61,7 @@
     columns=[
         "k",
         "distance2score",
-        "weight(sift vs color)",
+        "weight(sift vs color vs resnet)",
         "map@1",
         "map@5",
         "map@10",
@@ -82,10 +82,10 @@
 siftbow = SIFTBOWExtractor(mode="tfidf")
 sift_array_store = NPArrayStore(retrieve=KNNRetrieval(metric="manhattan"))
 
-rgb_histogram = RGBHistogram(n_bin=4, h_type="region")
+rgb_histogram = RGBHistogram(n_bin=4, h_type="region", n_slice=5)
 color_array_store = NPArrayStore(retrieve=KNNRetrieval(metric="cosine"))
 
-resnet = ResNetExtractor(model="resnet152", device="cuda")
+resnet = ResNetExtractor(model="resnet101", device="cuda")
 resnet_array_store = NPArrayStore(retrieve=KNNRetrieval(metric="cosine"))
 # Fitting siftbow with train data
 print("Fitting BOW for n_clusters kmeans: ", 96)
@@ -178,6 +178,7 @@
         for i in r:
             predicted.append(i.index)
         class_preds = np.take(dataset.targets, predicted, axis=0)
+        predicted = np.array(predicted).tolist()
         ap1.append(average_precision(class_preds.tolist(), [g.tolist()], 1))
         hit1.append(hit_rate(class_preds.tolist(), [g.tolist()], 1))
         recall1.append(recall(predicted, np.where(np.isin(np.array(dataset.targets), [g.tolist()]))[0], 1))

ensemble_sift_color.py

@@ -82,7 +82,7 @@
 siftbow = SIFTBOWExtractor(mode="tfidf")
 sift_array_store = NPArrayStore(retrieve=KNNRetrieval(metric="manhattan"))
 
-rgb_histogram = RGBHistogram(n_bin=4, h_type="region")
+rgb_histogram = RGBHistogram(n_bin=4, h_type="region", n_slice=5)
 color_array_store = NPArrayStore(retrieve=KNNRetrieval(metric="cosine"))
 
 # Fitting siftbow with train data

evaluation.sh

@@ -1,5 +1,5 @@
 # TQDM_DISABLE=1 python histogram_knn_eval.py
 # TQDM_DISABLE=1 python sift_knn_eval.py
 # TQDM_DISABLE=1 python resnet_knn_gpu_eval.py
-TQDM_DISABLE=1 python ensemble_sift_color.py
+# TQDM_DISABLE=1 python ensemble_sift_color.py
 TQDM_DISABLE=1 python ensemble_all.py

histogram_knn_eval.py

@@ -61,6 +61,7 @@
     columns=[
         "bins",
         "htype",
+        "slice",
         "metric",
         "map@1",
         "map@5",
@@ -80,12 +81,13 @@
 
 n_bins = [2, 4, 8, 12]
 h_types = ["region", "global"]
+n_slices = [2, 3, 4, 5]
 knn_metrics = ["euclidean", "cosine"]
-for bins, h_type, metric in grid(n_bins, h_types, knn_metrics):
-    print("Evaluate for bins: ", bins, " h_type: ", h_type, "", " with knn metric: ", metric)
+for bins, h_type, n_slice, metric in grid(n_bins, h_types, n_slices, knn_metrics):
+    print("Evaluate for bins: ", bins, " h_type: ", h_type, " n_slice: ", n_slice, " with knn metric: ", metric)
 
     # Initialization
-    rgb_histogram = RGBHistogram(n_bin=bins, h_type=h_type)
+    rgb_histogram = RGBHistogram(n_bin=bins, n_slice=n_slice, h_type=h_type)
     array_store = NPArrayStore(retrieve=KNNRetrieval(metric=metric))
     cbir = CBIR(rgb_histogram, array_store)
 
@@ -152,6 +154,7 @@
         {
             "bins": [bins],
             "htype": [h_type],
+            "slice": [n_slice],
             "metric": [metric],
             "map@1": [map1],
             "map@5": [map5],

out/ensemble_sift_color_resnet_knn_eval.csv

@@ -1,6 +1,6 @@
-k,distance2score,weight(sift vs color),map@1,map@5,map@10,hit_rate@1,hit_rate@5,hit_rate@10,recall@1,recall@5,recall@10,recall@100,recall@1000,avg_indexing_time,avg_retrieval_time,weight(sift vs color vs resnet)
-6353,exp,,0.879863,0.827304,0.795017,0.879863,0.965188,0.974061,0.0,0.0,0.0,0.0,0.0,0.012997,0.043442,"(0.16, 0.04, 0.8)"
-6353,exp,,0.869625,0.814334,0.780819,0.869625,0.956997,0.973379,0.0,0.0,0.0,0.0,0.0,0.012997,0.003675,"(0.32, 0.08, 0.6)"
-6353,exp,,0.85802,0.803413,0.764642,0.85802,0.952901,0.969283,0.0,0.0,0.0,0.0,0.0,0.012997,0.004133,"(0.4, 0.1, 0.5)"
-6353,exp,,0.838225,0.778703,0.736997,0.838225,0.944027,0.965188,0.0,0.0,0.0,0.0,0.0,0.012997,0.004157,"(0.48, 0.12, 0.4)"
-6353,exp,,0.745392,0.672765,0.62157,0.745392,0.884642,0.924232,0.0,0.0,0.0,0.0,0.0,0.012997,0.003683,"(0.64, 0.16, 0.2)"
+k,distance2score,weight(sift vs color vs resnet),map@1,map@5,map@10,hit_rate@1,hit_rate@5,hit_rate@10,recall@1,recall@5,recall@10,recall@100,recall@1000,avg_indexing_time,avg_retrieval_time
+6353,exp,"(0.16, 0.04, 0.8)",0.436519,0.722009,0.774087,0.873038,0.838771,0.801843,0.014099,0.066712,0.125761,0.557948,0.929494,0.012309,0.048446
+6353,exp,"(0.32, 0.08, 0.6)",0.433788,0.716522,0.766847,0.867577,0.826485,0.790034,0.013908,0.065267,0.122725,0.54261,0.921546,0.012309,0.003712
+6353,exp,"(0.4, 0.1, 0.5)",0.430375,0.707799,0.756372,0.860751,0.813925,0.77802,0.013712,0.063743,0.119679,0.523309,0.910957,0.012309,0.00409
+6353,exp,"(0.48, 0.12, 0.4)",0.422867,0.693414,0.739999,0.845734,0.790853,0.7557,0.013352,0.061367,0.114479,0.488053,0.889254,0.012309,0.00414
+6353,exp,"(0.64, 0.16, 0.2)",0.379181,0.619229,0.65654,0.758362,0.689283,0.636792,0.011448,0.049881,0.087719,0.34638,0.756345,0.012309,0.004134

out/histogram_knn_eval.csv

@@ -1,17 +1,65 @@
-bins,htype,metric,map@1,map@5,map@10,hit_rate@1,hit_rate@5,hit_rate@10,recall@1,recall@5,recall@10,recall@100,recall@1000,avg_indexing_time,avg_retrieval_time
-2,region,euclidean,0.331058,0.267986,0.239795,0.331058,0.483276,0.550853,0.003012,0.010598,0.017388,0.078739,0.310192,0.000992,0.000971
-2,region,cosine,0.324232,0.26744,0.239454,0.324232,0.475768,0.554949,0.00294,0.01054,0.017698,0.080455,0.312565,0.000998,0.001458
-2,global,euclidean,0.141297,0.106758,0.099454,0.141297,0.288737,0.382935,0.001118,0.003906,0.007064,0.044466,0.252985,0.000994,0.000713
-2,global,cosine,0.138567,0.107167,0.098157,0.138567,0.296246,0.386348,0.00113,0.003966,0.007086,0.044571,0.254884,0.000988,0.001246
-4,region,euclidean,0.369966,0.292014,0.262184,0.369966,0.498294,0.584983,0.003627,0.011819,0.019838,0.083359,0.318225,0.001009,0.001027
-4,region,cosine,0.373379,0.298976,0.263481,0.373379,0.509215,0.578157,0.003735,0.012632,0.020159,0.088442,0.322851,0.000988,0.003767
-4,global,euclidean,0.243686,0.194266,0.175973,0.243686,0.411604,0.488055,0.001998,0.007104,0.01214,0.06312,0.282745,0.000986,0.000991
-4,global,cosine,0.238908,0.193993,0.175563,0.238908,0.417747,0.486007,0.002068,0.007408,0.012499,0.063244,0.288249,0.000996,0.001339
-8,region,euclidean,0.352901,0.273311,0.23884,0.352901,0.483276,0.548123,0.003375,0.011001,0.017678,0.076686,0.308355,0.001082,0.002736
-8,region,cosine,0.356997,0.279317,0.248601,0.356997,0.486007,0.546758,0.003496,0.011554,0.018921,0.084385,0.325361,0.001062,0.025872
-8,global,euclidean,0.275085,0.216246,0.189283,0.275085,0.427304,0.48942,0.002263,0.007664,0.012529,0.061587,0.292441,0.001007,0.002483
-8,global,cosine,0.269625,0.207918,0.182935,0.269625,0.417065,0.496246,0.002455,0.00777,0.012896,0.063367,0.297811,0.001001,0.002484
-12,region,euclidean,0.327645,0.254334,0.221365,0.327645,0.458703,0.529693,0.00304,0.0106,0.017018,0.073402,0.298012,0.001484,0.007101
-12,region,cosine,0.338567,0.272901,0.239181,0.338567,0.479181,0.543345,0.003345,0.011587,0.018738,0.083657,0.322693,0.001499,0.067846
-12,global,euclidean,0.285324,0.215017,0.190512,0.285324,0.410239,0.487372,0.002313,0.007455,0.012528,0.061049,0.296934,0.001011,0.005967
-12,global,cosine,0.278498,0.208464,0.181229,0.278498,0.410239,0.481911,0.002527,0.007385,0.01234,0.062641,0.300355,0.000999,0.007284
+bins,htype,n_slice,metric,map@1,map@5,map@10,hit_rate@1,hit_rate@5,hit_rate@10,recall@1,recall@5,recall@10,recall@100,recall@1000,avg_indexing_time,avg_retrieval_time
+2,region,2,euclidean,0.106826,0.173152,0.18433,0.213652,0.165597,0.149488,0.001938,0.006562,0.010754,0.05622,0.26924,0.000879,0.000923
+2,region,2,cosine,0.102048,0.170466,0.18241,0.204096,0.164369,0.149147,0.001868,0.006417,0.011055,0.056983,0.272556,0.000866,0.000817
+2,region,3,euclidean,0.16587,0.264355,0.278764,0.331741,0.26744,0.238976,0.003029,0.010494,0.017425,0.078481,0.310125,0.001258,0.004554
+2,region,3,cosine,0.16314,0.261268,0.276275,0.32628,0.266075,0.238498,0.002953,0.010534,0.017511,0.08071,0.31222,0.00089,0.000947
+2,region,4,euclidean,0.173379,0.28108,0.296014,0.346758,0.288737,0.257474,0.003234,0.011733,0.019019,0.083794,0.319931,0.00093,0.003416
+2,region,4,cosine,0.172014,0.281812,0.298373,0.344027,0.291877,0.260137,0.003285,0.012027,0.019646,0.084507,0.320707,0.001403,0.005181
+2,region,5,euclidean,0.181911,0.292352,0.308463,0.363823,0.30116,0.266894,0.003433,0.012313,0.019938,0.086442,0.325511,0.001429,0.00333
+2,region,5,cosine,0.186348,0.295673,0.311676,0.372696,0.302935,0.272355,0.00364,0.012746,0.021034,0.087252,0.323152,0.001085,0.00535
+2,global,2,euclidean,0.071672,0.115802,0.125626,0.143345,0.106758,0.100205,0.001211,0.003891,0.00713,0.044607,0.253462,0.001433,0.000895
+2,global,2,cosine,0.066894,0.114906,0.124947,0.133788,0.108669,0.098635,0.001114,0.004044,0.007102,0.044757,0.255004,0.000957,0.000796
+2,global,3,euclidean,0.071672,0.115802,0.125626,0.143345,0.106758,0.100205,0.001211,0.003891,0.00713,0.044607,0.253462,0.000952,0.000726
+2,global,3,cosine,0.066894,0.114906,0.124947,0.133788,0.108669,0.098635,0.001114,0.004044,0.007102,0.044757,0.255004,0.000949,0.00079
+2,global,4,euclidean,0.071672,0.115802,0.125626,0.143345,0.106758,0.100205,0.001211,0.003891,0.00713,0.044607,0.253462,0.001182,0.000876
+2,global,4,cosine,0.066894,0.114906,0.124947,0.133788,0.108669,0.098635,0.001114,0.004044,0.007102,0.044757,0.255004,0.001169,0.000936
+2,global,5,euclidean,0.071672,0.115802,0.125626,0.143345,0.106758,0.100205,0.001211,0.003891,0.00713,0.044607,0.253462,0.001175,0.000925
+2,global,5,cosine,0.066894,0.114906,0.124947,0.133788,0.108669,0.098635,0.001114,0.004044,0.007102,0.044757,0.255004,0.001155,0.000952
+4,region,2,euclidean,0.156314,0.241024,0.254166,0.312628,0.238635,0.213584,0.002866,0.009144,0.015148,0.067827,0.29085,0.001166,0.00521
+4,region,2,cosine,0.151195,0.238561,0.251629,0.302389,0.236041,0.20901,0.002689,0.009304,0.015255,0.070237,0.29478,0.000878,0.003411
+4,region,3,euclidean,0.183276,0.288435,0.303298,0.366553,0.294334,0.263549,0.003521,0.011923,0.019939,0.083408,0.317844,0.000948,0.001089
+4,region,3,cosine,0.186348,0.29344,0.307859,0.372696,0.300341,0.264778,0.003696,0.012621,0.02028,0.088461,0.323241,0.00094,0.005043
+4,region,4,euclidean,0.187713,0.294155,0.310326,0.375427,0.300478,0.269147,0.003728,0.012061,0.020195,0.085404,0.326883,0.001,0.001219
+4,region,4,cosine,0.191126,0.303041,0.318549,0.382253,0.311126,0.277884,0.003869,0.012984,0.0215,0.091159,0.333826,0.001348,0.011747
+4,region,5,euclidean,0.190785,0.299006,0.314577,0.38157,0.304437,0.274061,0.003686,0.012554,0.020947,0.087145,0.329653,0.001466,0.008605
+4,region,5,cosine,0.198294,0.311802,0.328849,0.396587,0.319044,0.283481,0.004002,0.013556,0.022234,0.09315,0.337464,0.001409,0.012869
+4,global,2,euclidean,0.121843,0.199067,0.210732,0.243686,0.195495,0.175973,0.002061,0.007142,0.012122,0.06316,0.28265,0.00089,0.001038
+4,global,2,cosine,0.120137,0.198503,0.211282,0.240273,0.193993,0.175563,0.002134,0.007359,0.012559,0.063193,0.288359,0.000884,0.000842
+4,global,3,euclidean,0.121843,0.199067,0.210732,0.243686,0.195495,0.175973,0.002061,0.007142,0.012122,0.06316,0.28265,0.000884,0.003273
+4,global,3,cosine,0.120137,0.198503,0.211282,0.240273,0.193993,0.175563,0.002134,0.007359,0.012559,0.063193,0.288359,0.001179,0.001081
+4,global,4,euclidean,0.121843,0.199067,0.210732,0.243686,0.195495,0.175973,0.002061,0.007142,0.012122,0.06316,0.28265,0.001179,0.006888
+4,global,4,cosine,0.120137,0.198503,0.211282,0.240273,0.193993,0.175563,0.002134,0.007359,0.012559,0.063193,0.288359,0.001135,0.001096
+4,global,5,euclidean,0.121843,0.199067,0.210732,0.243686,0.195495,0.175973,0.002061,0.007142,0.012122,0.06316,0.28265,0.00097,0.001047
+4,global,5,cosine,0.120137,0.198503,0.211282,0.240273,0.193993,0.175563,0.002134,0.007359,0.012559,0.063193,0.288359,0.000893,0.00083
+8,region,2,euclidean,0.15802,0.242164,0.252948,0.316041,0.233038,0.206348,0.002934,0.008869,0.014289,0.064734,0.292216,0.000961,0.001677
+8,region,2,cosine,0.153925,0.234716,0.245759,0.30785,0.226348,0.197747,0.002963,0.009102,0.014268,0.06829,0.30101,0.000962,0.016444
+8,region,3,euclidean,0.175085,0.273295,0.284824,0.350171,0.274266,0.23884,0.003292,0.011137,0.017778,0.077265,0.30839,0.001452,0.010769
+8,region,3,cosine,0.180887,0.278588,0.291229,0.361775,0.280683,0.248191,0.003525,0.011633,0.018993,0.084323,0.325748,0.001424,0.028089
+8,region,4,euclidean,0.172696,0.267865,0.281322,0.345392,0.266485,0.237543,0.003317,0.010827,0.017912,0.076892,0.310431,0.001609,0.015663
+8,region,4,cosine,0.184642,0.286573,0.301817,0.369283,0.288874,0.259727,0.003563,0.011691,0.019297,0.087151,0.332759,0.001515,0.043942
+8,region,5,euclidean,0.174744,0.2688,0.281835,0.349488,0.268942,0.234744,0.003373,0.011235,0.018235,0.076498,0.310709,0.001614,0.006288
+8,region,5,cosine,0.191126,0.296932,0.31225,0.382253,0.301297,0.268942,0.003784,0.012575,0.020292,0.089297,0.336926,0.001542,0.062881
+8,global,2,euclidean,0.138908,0.219745,0.230661,0.277816,0.216928,0.190922,0.002308,0.00765,0.012624,0.061646,0.292162,0.001217,0.010844
+8,global,2,cosine,0.135836,0.214062,0.226496,0.271672,0.207509,0.18273,0.002467,0.007679,0.012749,0.063429,0.298087,0.000887,0.003786
+8,global,3,euclidean,0.138908,0.219745,0.230661,0.277816,0.216928,0.190922,0.002308,0.00765,0.012624,0.061646,0.292162,0.000895,0.002508
+8,global,3,cosine,0.135836,0.214062,0.226496,0.271672,0.207509,0.18273,0.002467,0.007679,0.012749,0.063429,0.298087,0.000878,0.00378
+8,global,4,euclidean,0.138908,0.219745,0.230661,0.277816,0.216928,0.190922,0.002308,0.00765,0.012624,0.061646,0.292162,0.000882,0.002841
+8,global,4,cosine,0.135836,0.214062,0.226496,0.271672,0.207509,0.18273,0.002467,0.007679,0.012749,0.063429,0.298087,0.001156,0.005243
+8,global,5,euclidean,0.138908,0.219745,0.230661,0.277816,0.216928,0.190922,0.002308,0.00765,0.012624,0.061646,0.292162,0.001074,0.012414
+8,global,5,cosine,0.135836,0.214062,0.226496,0.271672,0.207509,0.18273,0.002467,0.007679,0.012749,0.063429,0.298087,0.001154,0.004012
+12,region,2,euclidean,0.151195,0.232836,0.243736,0.302389,0.226894,0.1957,0.002794,0.008685,0.013843,0.063153,0.290062,0.001148,0.003813
+12,region,2,cosine,0.152218,0.233297,0.243683,0.304437,0.225666,0.194198,0.002989,0.009063,0.014293,0.067491,0.302006,0.001152,0.040478
+12,region,3,euclidean,0.162457,0.257606,0.268892,0.324915,0.256382,0.221092,0.003018,0.010735,0.017057,0.073701,0.297732,0.001567,0.00738
+12,region,3,cosine,0.17099,0.270285,0.283994,0.34198,0.273311,0.240887,0.00341,0.011592,0.018967,0.083543,0.32313,0.001556,0.07358
+12,region,4,euclidean,0.161433,0.249721,0.260406,0.322867,0.245051,0.210307,0.003072,0.010501,0.016416,0.072419,0.29958,0.002318,0.017848
+12,region,4,cosine,0.175427,0.28082,0.295944,0.350853,0.287372,0.25372,0.003373,0.011955,0.019279,0.087612,0.329533,0.002034,0.124746
+12,region,5,euclidean,0.157679,0.243503,0.254687,0.315358,0.236997,0.204164,0.003107,0.0105,0.016837,0.071375,0.298628,0.004012,0.026965
+12,region,5,cosine,0.181911,0.290048,0.305795,0.363823,0.298294,0.262526,0.003565,0.012416,0.020245,0.089304,0.334704,0.007279,0.191561
+12,global,2,euclidean,0.143686,0.220971,0.231971,0.287372,0.214198,0.189078,0.002316,0.007353,0.012357,0.060828,0.296786,0.000905,0.006021
+12,global,2,cosine,0.139932,0.217328,0.227818,0.279863,0.210648,0.182867,0.002527,0.007455,0.012323,0.062611,0.300949,0.000887,0.008011
+12,global,3,euclidean,0.143686,0.220971,0.231971,0.287372,0.214198,0.189078,0.002316,0.007353,0.012357,0.060828,0.296786,0.000904,0.005986
+12,global,3,cosine,0.139932,0.217328,0.227818,0.279863,0.210648,0.182867,0.002527,0.007455,0.012323,0.062611,0.300949,0.000902,0.008028
+12,global,4,euclidean,0.143686,0.220971,0.231971,0.287372,0.214198,0.189078,0.002316,0.007353,0.012357,0.060828,0.296786,0.000911,0.005983
+12,global,4,cosine,0.139932,0.217328,0.227818,0.279863,0.210648,0.182867,0.002527,0.007455,0.012323,0.062611,0.300949,0.000906,0.008035
+12,global,5,euclidean,0.143686,0.220971,0.231971,0.287372,0.214198,0.189078,0.002316,0.007353,0.012357,0.060828,0.296786,0.000897,0.006013
+12,global,5,cosine,0.139932,0.217328,0.227818,0.279863,0.210648,0.182867,0.002527,0.007455,0.012323,0.062611,0.300949,0.000904,0.008067

out/resnet_knn_cpu_eval.csv

@@ -0,0 +1,11 @@
+model,metric,map@1,map@5,map@10,hit_rate@1,hit_rate@5,hit_rate@10,recall@1,recall@5,recall@10,recall@100,recall@1000,avg_indexing_time,avg_retrieval_time
+resnet18,euclidean,0.399659,0.655243,0.699318,0.799317,0.745256,0.704573,0.012291,0.055233,0.101115,0.424731,0.819271,0.007408,0.003459
+resnet18,cosine,0.415017,0.678607,0.724738,0.830034,0.773925,0.73802,0.012926,0.058735,0.109174,0.473221,0.876809,0.007412,0.005232
+resnet34,euclidean,0.419454,0.689021,0.736365,0.838908,0.787713,0.750307,0.013275,0.060561,0.113126,0.485451,0.867174,0.011084,0.004421
+resnet34,cosine,0.426962,0.704586,0.754328,0.853925,0.813379,0.777884,0.013635,0.063591,0.119375,0.532325,0.9167,0.01154,0.005935
+resnet50,euclidean,0.422184,0.688405,0.73499,0.844369,0.786075,0.750785,0.013318,0.059945,0.112337,0.484026,0.87145,0.026898,0.009914
+resnet50,cosine,0.432082,0.70548,0.75444,0.864164,0.808055,0.774812,0.013846,0.062628,0.117531,0.525654,0.910517,0.027144,0.019858
+resnet101,euclidean,0.433788,0.709703,0.759503,0.867577,0.817201,0.784027,0.013994,0.064548,0.122043,0.524385,0.895737,0.038445,0.012667
+resnet101,cosine,0.437543,0.72387,0.775381,0.875085,0.838362,0.803891,0.014133,0.066707,0.126186,0.560837,0.931262,0.038086,0.022556
+resnet152,euclidean,0.432765,0.706929,0.754837,0.865529,0.80901,0.773311,0.013978,0.063604,0.119526,0.522956,0.899991,0.054021,0.016194
+resnet152,cosine,0.441297,0.721474,0.772377,0.882594,0.828123,0.797406,0.014217,0.065943,0.124771,0.562465,0.932063,0.053849,0.025914