WIP

annotations_filtered/q-1DREuXwjc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 2.24], [3.0, 2.83], [3.0, 7.64], [10.0, 10.2], [13.0, 13.76], [15.0, 18.27], [21.0, 21.07], [22.0, 22.17], [23.0, 23.45], [25.0, 26.55], [32.0, 32.41], [39.0, 39.73], [41.0, 41.45], [44.0, 44.51], [48.0, 48.95]], "keep_status": [false, false, true, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 38.84, 0.0, 0.0, 45.75, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, [["hum", 32.53], ["mains hum", 15.74], ["speech", 15.25]], null, null, [["speech", 40.42], ["sine wave", 30.86], ["chirp tone", 8.52]], null, null, null, null, null, null, null, null, null], "duration": [1.24, -0.17, 4.64, 0.2, 0.76, 3.27, 0.07, 0.17, 0.45, 1.55, 0.41, 0.73, 0.45, 0.51, 0.95]}

annotations_filtered/q-H62GgHjeg_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 4.43], [17.0, 17.58], [29.0, 30.48], [55.0, 55.76], [59.0, 60.18], [71.0, 72.47], [74.0, 75.73], [76.0, 77.6], [78.0, 78.22], [81.0, 81.01], [83.0, 83.62], [94.0, 94.78], [105.0, 105.85], [110.0, 114.35], [115.0, 115.37], [122.0, 123.42], [126.0, 126.52], [128.0, 128.97], [130.0, 133.51], [138.0, 137.86], [139.0, 145.69]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false, true], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 41.89, 0.0, 0.0, 0.0, 0.0, 32.84, 0.0, 31.65], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, [["fart", 34.47], ["speech", 29.46], ["music", 6.12]], null, null, null, null, [["cattle, bovinae", 23.0], ["music", 19.66], ["moo", 17.99]], null, [["speech", 26.02], ["cattle, bovinae", 24.3], ["moo", 17.48]]], "duration": [0.43, 0.58, 1.48, 0.76, 1.18, 1.47, 1.73, 1.6, 0.22, 0.01, 0.62, 0.78, 0.85, 4.35, 0.37, 1.42, 0.52, 0.97, 3.51, -0.14, 6.69]}

annotations_filtered/q-StMfE8NrA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 7.21], [8.0, 49.08], [51.0, 85.87]], "keep_status": [false, false, false], "silence_prob": [32.37, 0.0, 0.0], "audiomae_on_audioset": [[["hum", 34.28], ["music", 33.27], ["throbbing", 12.38]], null, null], "duration": [6.21, 41.08, 34.87]}

annotations_filtered/q-kLlfq4JpU_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[34.0, 33.89], [42.0, 42.69], [44.0, 43.78]], "keep_status": [false, false, false], "silence_prob": [0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null], "duration": [-0.11, 0.69, -0.22]}

annotations_filtered/q-nQtR-WbIs_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 10.99], [15.0, 16.09], [17.0, 18.67], [20.0, 21.05], [23.0, 24.83], [25.0, 27.5], [31.0, 31.35], [33.0, 33.89], [36.0, 42.15], [44.0, 45.4], [49.0, 53.2], [55.0, 57.52], [60.0, 61.84], [63.0, 65.65], [68.0, 70.68], [72.0, 91.34], [93.0, 97.53], [98.0, 101.28], [102.0, 109.75]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false], "silence_prob": [61.87, 0.0, 0.0, 0.0, 0.0, 60.42, 0.0, 0.0, 63.64, 0.0, 58.3, 80.11, 0.0, 73.51, 46.36, 53.22, 71.57, 53.4, 57.48], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 42.09], ["musical instrument", 12.16], ["guitar", 5.93]], null, null, null, null], "duration": [4.99, 1.09, 1.67, 1.05, 1.83, 2.5, 0.35, 0.89, 6.15, 1.4, 4.2, 2.52, 1.84, 2.65, 2.68, 19.34, 4.53, 3.28, 7.75]}

annotations_filtered/q-y6JBpCFtI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 3.1], [22.0, 23.14], [30.0, 31.24], [33.0, 34.59], [36.0, 37.62], [39.0, 40.53], [46.0, 47.65], [53.0, 53.43], [56.0, 57.03], [60.0, 63.61], [67.0, 68.25], [69.0, 70.36], [79.0, 80.13], [88.0, 99.45], [101.0, 101.01], [106.0, 106.32], [107.0, 107.69], [118.0, 120.75]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.62, 0.0, 0.0, 0.0, 97.83, 0.0, 0.0, 0.0, 93.13], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [1.1, 1.14, 1.24, 1.59, 1.62, 1.53, 1.65, 0.43, 1.03, 3.61, 1.25, 1.36, 1.13, 11.45, 0.01, 0.32, 0.69, 2.75]}

annotations_filtered/q06t8RTLqMQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 8.34], [10.0, 12.26], [15.0, 15.63], [17.0, 17.69], [25.0, 35.95], [36.0, 40.68], [41.0, 42.42], [43.0, 44.79], [47.0, 47.33], [50.0, 55.59], [57.0, 60.35], [62.0, 64.05], [66.0, 67.36], [71.0, 71.63], [74.0, 75.02], [83.0, 86.0], [88.0, 90.17], [92.0, 98.39], [102.0, 105.43], [107.0, 108.75], [110.0, 109.95], [113.0, 114.2], [115.0, 115.67]], "keep_status": [false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [92.64, 84.25, 0.0, 0.0, 33.39, 32.21, 0.0, 0.0, 0.0, 73.06, 40.99, 40.93, 0.0, 0.0, 0.0, 59.42, 46.61, 80.11, 41.28, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, [["dial tone", 47.99], ["speech", 28.52], ["sidetone", 8.24]], [["cattle, bovinae", 48.9], ["moo", 33.44], ["livestock, farm animals, working animals", 10.75]], null, null, null, null, [["dial tone", 19.85], ["chirp tone", 19.43], ["sine wave", 14.41]], [["cattle, bovinae", 38.64], ["moo", 36.11], ["livestock, farm animals, working animals", 4.39]], null, null, null, null, [["chirp tone", 53.41], ["sine wave", 17.39], ["busy signal", 3.8]], null, [["speech", 51.68], ["noise", 11.67], ["sidetone", 8.07]], null, null, null, null], "duration": [4.34, 2.26, 0.63, 0.69, 10.95, 4.68, 1.42, 1.79, 0.33, 5.59, 3.35, 2.05, 1.36, 0.63, 1.02, 3.0, 2.17, 6.39, 3.43, 1.75, -0.05, 1.2, 0.67]}

annotations_filtered/q0vvYHuA10U_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 10.32], [11.0, 21.81], [22.0, 21.9], [22.0, 39.26], [40.0, 46.63], [48.0, 56.4], [59.0, 76.94], [81.0, 85.7], [88.0, 91.1], [94.0, 106.15], [108.0, 108.85], [113.0, 117.41]], "keep_status": [true, true, false, true, true, false, true, true, true, false, false, false], "silence_prob": [29.39, 29.98, 0.0, 29.38, 29.85, 60.05, 32.6, 36.74, 33.44, 29.85, 0.0, 60.14], "audiomae_on_audioset": [[["speech", 56.75], ["explosion", 6.39], ["burst, pop", 4.82]], [["buzz", 37.02], ["hum", 11.18], ["music", 6.94]], null, [["music", 25.82], ["hum", 13.13], ["cacophony", 10.18]], [["speech", 57.7], ["vehicle", 4.2], ["explosion", 3.51]], null, [["music", 57.44], ["theremin", 9.18], ["electronic music", 2.88]], [["music", 46.44], ["brass instrument", 7.49], ["trombone", 4.7]], [["music", 37.71], ["speech", 15.08], ["didgeridoo", 4.84]], [["music", 50.91], ["buzz", 16.01], ["didgeridoo", 3.53]], null, null], "duration": [2.32, 10.81, -0.1, 17.26, 6.63, 8.4, 17.94, 4.7, 3.1, 12.15, 0.85, 4.41]}

annotations_filtered/q0yFqlPrLyE_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 7.43], [8.0, 15.52], [18.0, 18.47], [19.0, 22.17], [23.0, 29.94], [31.0, 39.31], [45.0, 52.56], [56.0, 61.28], [62.0, 63.34], [65.0, 65.8], [74.0, 74.58], [75.0, 78.41], [80.0, 80.28], [83.0, 82.9], [95.0, 95.32], [98.0, 99.87], [100.0, 100.79], [104.0, 105.78], [108.0, 109.17], [113.0, 113.9], [115.0, 115.42], [116.0, 117.24], [118.0, 119.79]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [99.85, 93.29, 0.0, 74.6, 41.98, 38.23, 35.42, 36.6, 0.0, 0.0, 0.0, 63.64, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, [["boing", 58.0], ["music", 28.85], ["speech", 6.57]], [["music", 75.79], ["musical instrument", 4.69], ["brass instrument", 2.0]], [["music", 66.78], ["speech", 7.5], ["livestock, farm animals, working animals", 3.17]], [["music", 82.57], ["singing", 1.21], ["musical instrument", 0.93]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [3.43, 7.52, 0.47, 3.17, 6.94, 8.31, 7.56, 5.28, 1.34, 0.8, 0.58, 3.41, 0.28, -0.1, 0.32, 1.87, 0.79, 1.78, 1.17, 0.9, 0.42, 1.24, 1.79]}

annotations_filtered/q15Yv0rZXqs_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 4.78], [9.0, 11.82], [14.0, 15.75], [36.0, 36.64], [41.0, 43.06], [46.0, 56.37], [60.0, 80.72], [83.0, 90.31], [92.0, 104.9], [108.0, 108.16]], "keep_status": [false, true, false, false, false, false, false, true, false, false], "silence_prob": [0.0, 46.12, 0.0, 0.0, 86.64, 43.85, 37.11, 35.41, 32.44, 0.0], "audiomae_on_audioset": [null, [["music", 52.11], ["brass instrument", 6.08], ["clarinet", 6.06]], null, null, null, [["speech", 52.36], ["music", 22.95], ["musical instrument", 5.35]], [["music", 69.08], ["didgeridoo", 16.59], ["musical instrument", 2.19]], [["music", 37.52], ["didgeridoo", 18.35], ["speech", 6.39]], [["cattle, bovinae", 41.85], ["moo", 30.65], ["livestock, farm animals, working animals", 14.96]], null], "duration": [1.78, 2.82, 1.75, 0.64, 2.06, 10.37, 20.72, 7.31, 12.9, 0.16]}

annotations_filtered/q1D9i-d1m4Y_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[7.0, 8.02], [26.0, 25.93], [38.0, 38.47], [43.0, 45.25], [55.0, 59.07], [61.0, 76.32]], "keep_status": [false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 44.96, 46.97, 43.74], "audiomae_on_audioset": [null, null, null, [["music", 92.01], ["speech", 0.89], ["musical instrument", 0.63]], [["music", 71.98], ["didgeridoo", 3.05], ["musical instrument", 2.5]], [["music", 76.88], ["trombone", 3.96], ["didgeridoo", 3.76]]], "duration": [1.02, -0.07, 0.47, 2.25, 4.07, 15.32]}

annotations_filtered/q1Pz7ppcuJc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 5.37], [6.0, 8.75], [9.0, 12.21], [13.0, 15.96], [20.0, 22.1], [23.0, 24.04], [26.0, 25.98], [28.0, 28.51], [29.0, 30.55], [32.0, 35.99], [37.0, 38.18], [40.0, 41.03], [44.0, 45.89], [47.0, 62.24], [63.0, 64.81], [66.0, 66.19], [67.0, 68.72], [69.0, 70.56], [74.0, 75.09], [77.0, 77.45], [78.0, 78.88], [80.0, 81.33], [83.0, 83.51], [87.0, 87.51], [88.0, 93.56], [96.0, 96.67], [97.0, 100.84], [103.0, 104.58], [105.0, 106.96], [111.0, 111.52], [112.0, 115.52], [117.0, 123.72], [129.0, 129.69]], "keep_status": [false, false, false, false, true, false, false, false, false, true, false, false, false, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [91.47, 84.98, 49.22, 85.72, 44.34, 0.0, 0.0, 0.0, 0.0, 48.06, 0.0, 0.0, 0.0, 41.01, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 50.71, 0.0, 51.82, 0.0, 0.0, 0.0, 54.04, 51.71, 0.0], "audiomae_on_audioset": [null, null, [["speech", 52.06], ["telephone", 13.12], ["sidetone", 7.11]], null, [["speech", 43.81], ["cat", 5.86], ["inside, small room", 3.66]], null, null, null, null, [["thunk", 15.77], ["noise", 14.97], ["crushing", 13.75]], null, null, null, [["music", 51.39], ["electric shaver, electric razor", 7.36], ["hum", 5.36]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [3.37, 2.75, 3.21, 2.96, 2.1, 1.04, -0.02, 0.51, 1.55, 3.99, 1.18, 1.03, 1.89, 15.24, 1.81, 0.19, 1.72, 1.56, 1.09, 0.45, 0.88, 1.33, 0.51, 0.51, 5.56, 0.67, 3.84, 1.58, 1.96, 0.52, 3.52, 6.72, 0.69]}

annotations_filtered/q1SFvQhjK5I_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[24.0, 41.28], [42.0, 42.26], [43.0, 43.68], [45.0, 45.15], [50.0, 50.8], [53.0, 54.06], [55.0, 56.54], [58.0, 58.8], [65.0, 65.67], [66.0, 67.54], [69.0, 69.52], [70.0, 71.07], [73.0, 73.72], [80.0, 81.82], [84.0, 84.94], [101.0, 101.95], [103.0, 104.45], [105.0, 106.68], [108.0, 109.07], [111.0, 111.54], [112.0, 112.97], [114.0, 114.64], [116.0, 118.1], [129.0, 130.01]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [37.73, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 97.43, 0.0], "audiomae_on_audioset": [[["music", 77.7], ["theremin", 2.9], ["throbbing", 1.33]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [17.28, 0.26, 0.68, 0.15, 0.8, 1.06, 1.54, 0.8, 0.67, 1.54, 0.52, 1.07, 0.72, 1.82, 0.94, 0.95, 1.45, 1.68, 1.07, 0.54, 0.97, 0.64, 2.1, 1.01]}

annotations_filtered/q1bV-D8cSz8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[15.0, 37.77], [39.0, 44.68], [49.0, 71.68], [72.0, 72.62], [77.0, 77.25], [79.0, 79.62], [81.0, 82.07], [83.0, 96.45], [96.0, 115.42]], "keep_status": [false, false, false, false, false, false, false, true, false], "silence_prob": [80.11, 86.82, 72.75, 0.0, 0.0, 0.0, 0.0, 32.66, 33.61], "audiomae_on_audioset": [null, null, null, null, null, null, null, [["music", 23.95], ["speech", 16.88], ["hum", 14.72]], [["music", 37.04], ["speech", 24.19], ["smash, crash", 9.74]]], "duration": [22.77, 5.68, 22.68, 0.62, 0.25, 0.62, 1.07, 13.45, 19.42]}

annotations_filtered/q254XDNZ2Ao_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[14.0, 30.75], [36.0, 38.52], [44.0, 48.05], [49.0, 50.36], [51.0, 85.09], [87.0, 116.95], [117.0, 117.02], [118.0, 117.78]], "keep_status": [true, true, true, false, false, true, false, false], "silence_prob": [31.56, 32.37, 39.41, 0.0, 0.0, 30.1, 0.0, 0.0], "audiomae_on_audioset": [[["music", 20.43], ["didgeridoo", 9.59], ["theremin", 6.52]], [["speech", 25.4], ["race car, auto racing", 8.76], ["vehicle", 5.75]], [["music", 34.53], ["fly, housefly", 7.43], ["mosquito", 5.22]], null, null, [["music", 15.18], ["fart", 14.86], ["whack, thwack", 10.16]], null, null], "duration": [16.75, 2.52, 4.05, 1.36, 34.09, 29.95, 0.02, -0.22]}

annotations_filtered/q289a8P8Ht8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[34.0, 116.88], [118.0, 149.82]], "keep_status": [false, false], "silence_prob": [0.0, 0.0], "audiomae_on_audioset": [null, null], "duration": [82.88, 31.82]}

annotations_filtered/q292IDwEWZ0_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 19.99], [22.0, 22.94], [24.0, 25.79], [27.0, 30.15], [31.0, 43.44], [44.0, 44.31], [44.0, 44.37], [44.0, 46.87], [50.0, 83.42], [86.0, 101.17], [102.0, 102.41], [109.0, 109.85], [110.0, 118.83], [121.0, 134.59]], "keep_status": [false, false, false, true, false, false, false, true, false, false, false, false, true, true], "silence_prob": [29.32, 0.0, 0.0, 29.09, 29.06, 0.0, 0.0, 29.17, 0.0, 31.86, 0.0, 0.0, 29.92, 28.79], "audiomae_on_audioset": [[["music", 38.93], ["speech", 35.7], ["thunk", 2.42]], null, null, [["theremin", 31.87], ["music", 27.32], ["crowd", 3.77]], [["cattle, bovinae", 33.73], ["moo", 20.29], ["livestock, farm animals, working animals", 18.62]], null, null, [["whack, thwack", 30.92], ["throbbing", 15.67], ["whip", 9.01]], null, [["music", 44.22], ["theremin", 21.93], ["ambient music", 5.12]], null, null, [["music", 22.35], ["fly, housefly", 22.3], ["insect", 7.58]], [["livestock, farm animals, working animals", 13.56], ["music", 12.15], ["cattle, bovinae", 6.67]]], "duration": [18.99, 0.94, 1.79, 3.15, 12.44, 0.31, 0.37, 2.87, 33.42, 15.17, 0.41, 0.85, 8.83, 13.59]}

annotations_filtered/q2EU-k9I5yg_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 3.48], [4.0, 4.45], [10.0, 10.39], [12.0, 12.04], [13.0, 13.61], [19.0, 19.57], [45.0, 45.6], [48.0, 48.61], [58.0, 59.19], [97.0, 97.12], [101.0, 101.33], [102.0, 102.1], [106.0, 107.32], [111.0, 111.91], [113.0, 113.64]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [92.64, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [2.48, 0.45, 0.39, 0.04, 0.61, 0.57, 0.6, 0.61, 1.19, 0.12, 0.33, 0.1, 1.32, 0.91, 0.64]}

annotations_filtered/q2YwvMc96VY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 3.3], [5.0, 6.02], [7.0, 7.05], [9.0, 9.81], [11.0, 11.53], [14.0, 14.57], [16.0, 16.44], [20.0, 20.17], [22.0, 64.52], [65.0, 67.37], [68.0, 69.6], [70.0, 72.82], [73.0, 75.36], [76.0, 77.63], [79.0, 79.44], [80.0, 80.76], [83.0, 83.24], [85.0, 86.07], [87.0, 90.54], [91.0, 100.33], [103.0, 104.14], [107.0, 108.8], [109.0, 111.92]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 72.9, 0.0, 57.48, 64.07, 0.0, 0.0, 0.0, 0.0, 0.0, 92.8, 33.63, 0.0, 0.0, 52.1], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 47.18], ["speech", 15.51], ["sidetone", 5.07]], null, null, null], "duration": [1.3, 1.02, 0.05, 0.81, 0.53, 0.57, 0.44, 0.17, 42.52, 2.37, 1.6, 2.82, 2.36, 1.63, 0.44, 0.76, 0.24, 1.07, 3.54, 9.33, 1.14, 1.8, 2.92]}

annotations_filtered/q2pzOimT9so_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 26.87], [27.0, 27.67], [29.0, 32.16], [34.0, 43.5], [44.0, 48.93], [51.0, 52.79], [55.0, 57.23], [58.0, 63.71], [68.0, 69.31]], "keep_status": [false, false, false, false, false, false, false, false, false], "silence_prob": [73.97, 0.0, 96.42, 99.97, 97.11, 0.0, 84.43, 99.44, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null], "duration": [18.87, 0.67, 3.16, 9.5, 4.93, 1.79, 2.23, 5.71, 1.31]}

annotations_filtered/q30Pl1M6_DE_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[18.0, 110.42], [113.0, 112.9], [114.0, 136.07]], "keep_status": [false, false, false], "silence_prob": [0.0, 0.0, 29.54], "audiomae_on_audioset": [null, null, [["music", 69.57], ["hum", 9.74], ["mains hum", 4.53]]], "duration": [92.42, -0.1, 22.07]}

annotations_filtered/q3JlGPF4Ko8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 3.06], [5.0, 5.68], [6.0, 8.14], [10.0, 10.89], [12.0, 13.42], [14.0, 18.84], [22.0, 29.89], [34.0, 35.33], [39.0, 47.58], [49.0, 50.01], [52.0, 53.5], [56.0, 58.6], [61.0, 63.51], [65.0, 65.89], [72.0, 115.67], [117.0, 118.32], [120.0, 121.95]], "keep_status": [true, false, true, false, false, false, false, false, true, false, false, true, true, false, false, false, false], "silence_prob": [30.41, 0.0, 39.24, 0.0, 0.0, 42.3, 54.1, 0.0, 41.64, 0.0, 0.0, 49.36, 35.68, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [[["sound effect", 10.76], ["music", 9.96], ["trombone", 8.21]], null, [["hum", 22.74], ["sidetone", 20.33], ["mains hum", 19.7]], null, null, [["speech", 46.75], ["radio", 26.61], ["hum", 4.39]], null, null, [["hum", 26.72], ["mains hum", 25.87], ["vehicle", 8.24]], null, null, [["hum", 6.52], ["vehicle", 6.36], ["creak", 5.51]], [["speech", 18.55], ["radio", 11.12], ["frog", 7.68]], null, null, null, null], "duration": [2.06, 0.68, 2.14, 0.89, 1.42, 4.84, 7.89, 1.33, 8.58, 1.01, 1.5, 2.6, 2.51, 0.89, 43.67, 1.32, 1.95]}

annotations_filtered/q3NI5sE3KeY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[10.0, 10.23], [16.0, 17.61], [22.0, 22.77], [26.0, 26.47], [30.0, 30.91], [41.0, 41.64], [58.0, 59.07], [60.0, 61.59], [63.0, 64.66], [78.0, 78.58], [80.0, 80.11], [83.0, 90.98], [93.0, 94.41], [96.0, 97.39], [102.0, 104.58], [106.0, 107.2], [112.0, 112.77], [117.0, 117.04], [122.0, 122.91], [127.0, 126.86], [132.0, 134.25], [138.0, 137.74], [143.0, 143.67], [158.0, 159.87], [164.0, 164.25], [181.0, 181.28]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 83.88, 0.0, 0.0, 56.03, 0.0, 0.0, 0.0, 0.0, 0.0, 48.35, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["didgeridoo", 46.53], ["music", 29.43], ["mains hum", 4.95]], null, null, null, null, null], "duration": [0.23, 1.61, 0.77, 0.47, 0.91, 0.64, 1.07, 1.59, 1.66, 0.58, 0.11, 7.98, 1.41, 1.39, 2.58, 1.2, 0.77, 0.04, 0.91, -0.14, 2.25, -0.26, 0.67, 1.87, 0.25, 0.28]}

annotations_filtered/q3OTEdZkBaQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[12.0, 15.7], [27.0, 31.75], [33.0, 35.41], [36.0, 37.18], [53.0, 54.16], [56.0, 56.61], [70.0, 70.33], [73.0, 78.49], [80.0, 80.57], [82.0, 82.32]], "keep_status": [true, false, false, false, false, false, false, true, false, false], "silence_prob": [32.69, 34.04, 32.52, 0.0, 0.0, 0.0, 0.0, 30.5, 0.0, 0.0], "audiomae_on_audioset": [[["speech", 25.18], ["fly, housefly", 17.96], ["music", 13.51]], [["hum", 43.48], ["mains hum", 25.16], ["whale vocalization", 7.12]], [["mains hum", 32.88], ["whale vocalization", 23.9], ["hum", 17.33]], null, null, null, null, [["music", 41.53], ["speech", 16.67], ["fart", 6.5]], null, null], "duration": [3.7, 4.75, 2.41, 1.18, 1.16, 0.61, 0.33, 5.49, 0.57, 0.32]}

annotations_filtered/q3Vvto0REuc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[21.0, 22.05], [28.0, 39.77], [41.0, 41.91], [44.0, 46.67], [48.0, 50.04], [51.0, 52.05], [53.0, 54.06], [56.0, 56.49], [57.0, 58.23], [63.0, 63.83], [65.0, 65.74], [67.0, 70.12], [71.0, 72.1], [73.0, 73.99], [75.0, 75.95], [77.0, 77.53], [79.0, 79.88], [81.0, 82.31], [83.0, 83.47], [85.0, 85.77], [88.0, 91.86], [93.0, 94.86], [97.0, 97.16], [99.0, 99.03], [101.0, 101.61], [102.0, 103.25], [104.0, 104.38], [105.0, 104.94], [107.0, 107.76], [109.0, 109.93], [111.0, 112.24], [115.0, 115.08], [116.0, 117.42], [119.0, 119.57], [122.0, 123.16], [124.0, 124.78], [126.0, 126.69], [128.0, 128.73], [130.0, 130.98], [133.0, 133.32], [134.0, 135.01], [136.0, 137.03], [139.0, 143.16], [148.0, 149.88], [150.0, 152.56], [153.0, 154.75], [156.0, 157.17], [158.0, 160.1], [161.0, 162.08], [164.0, 166.53], [168.0, 168.89], [172.0, 193.24], [194.0, 195.61], [197.0, 197.46], [200.0, 202.95], [204.0, 205.43], [208.0, 215.89], [217.0, 219.06], [219.0, 221.97], [222.0, 224.14], [225.0, 228.33]], "keep_status": [false, true, false, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false], "silence_prob": [0.0, 30.57, 0.0, 49.68, 98.93, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 98.86, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.92, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.91, 0.0, 56.18, 0.0, 0.0, 53.84, 0.0, 99.76, 0.0, 94.95, 0.0, 0.0, 78.55, 0.0, 99.88, 62.78, 49.27, 39.63, 99.16], "audiomae_on_audioset": [null, [["fly, housefly", 15.2], ["hum", 15.1], ["music", 13.9]], null, [["speech", 11.13], ["hum", 10.58], ["sneeze", 9.63]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["speech", 49.07], ["sidetone", 24.02], ["hum", 3.63]], [["speech", 31.75], ["sidetone", 19.56], ["dial tone", 12.14]], null], "duration": [1.05, 11.77, 0.91, 2.67, 2.04, 1.05, 1.06, 0.49, 1.23, 0.83, 0.74, 3.12, 1.1, 0.99, 0.95, 0.53, 0.88, 1.31, 0.47, 0.77, 3.86, 1.86, 0.16, 0.03, 0.61, 1.25, 0.38, -0.06, 0.76, 0.93, 1.24, 0.08, 1.42, 0.57, 1.16, 0.78, 0.69, 0.73, 0.98, 0.32, 1.01, 1.03, 4.16, 1.88, 2.56, 1.75, 1.17, 2.1, 1.08, 2.53, 0.89, 21.24, 1.61, 0.46, 2.95, 1.43, 7.89, 2.06, 2.97, 2.14, 3.33]}

annotations_filtered/q3a5wxfm13Q_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 5.66], [7.0, 10.94], [12.0, 11.87], [17.0, 17.61], [18.0, 18.67], [20.0, 20.92], [22.0, 22.28], [26.0, 26.65], [31.0, 31.13], [37.0, 36.85], [37.0, 38.23], [40.0, 40.58], [42.0, 42.21], [43.0, 43.39], [46.0, 46.75], [49.0, 49.17], [52.0, 52.03], [57.0, 56.86], [58.0, 58.75], [63.0, 62.99], [65.0, 65.92], [68.0, 70.95], [71.0, 71.71], [72.0, 72.6], [83.0, 83.02], [85.0, 98.31], [99.0, 99.4], [101.0, 101.46], [103.0, 107.49], [109.0, 111.91]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 84.98, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 32.56, 0.0, 0.0, 0.0, 91.13, 0.0, 0.0, 31.23, 57.56], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["speech", 82.45], ["radio", 10.48], ["speech synthesizer", 1.65]], null, null, null, null, null, null, [["speech", 84.17], ["radio", 8.92], ["speech synthesizer", 1.88]], null], "duration": [0.66, 3.94, -0.13, 0.61, 0.67, 0.92, 0.28, 0.65, 0.13, -0.15, 1.23, 0.58, 0.21, 0.39, 0.75, 0.17, 0.03, -0.14, 0.75, -0.01, 0.92, 2.95, 0.71, 0.6, 0.02, 13.31, 0.4, 0.46, 4.49, 2.91]}

annotations_filtered/q42thgSKkpo_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 64.52], [67.0, 75.03], [76.0, 81.19], [83.0, 83.44], [84.0, 83.61], [84.0, 99.37], [103.0, 110.79], [116.0, 118.81], [122.0, 132.19], [133.0, 153.87], [155.0, 158.97]], "keep_status": [false, true, true, false, false, false, true, false, false, false, true], "silence_prob": [0.0, 29.06, 30.16, 0.0, 0.0, 29.37, 42.19, 36.52, 35.61, 38.9, 29.26], "audiomae_on_audioset": [null, [["music", 26.49], ["speech", 23.9], ["mains hum", 13.67]], [["music", 25.8], ["mains hum", 6.93], ["foghorn", 6.43]], null, null, [["music", 65.23], ["didgeridoo", 7.31], ["speech", 3.64]], [["music", 41.02], ["speech", 15.76], ["whale vocalization", 7.68]], [["theremin", 36.28], ["music", 31.47], ["wind instrument, woodwind instrument", 4.54]], [["music", 46.12], ["theremin", 25.52], ["whale vocalization", 4.7]], [["didgeridoo", 43.17], ["music", 34.13], ["musical instrument", 2.84]], [["music", 45.52], ["musical instrument", 9.54], ["sound effect", 6.44]]], "duration": [56.52, 8.03, 5.19, 0.44, -0.39, 15.37, 7.79, 2.81, 10.19, 20.87, 3.97]}

annotations_filtered/q437KEcmwmM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[33.0, 37.13], [43.0, 43.48], [44.0, 49.01], [52.0, 85.55], [88.0, 93.88], [95.0, 96.89], [97.0, 104.48]], "keep_status": [true, false, true, false, false, false, false], "silence_prob": [35.0, 0.0, 30.58, 0.0, 70.72, 0.0, 46.05], "audiomae_on_audioset": [[["music", 9.82], ["vehicle", 6.77], ["eruption", 5.16]], null, [["music", 34.67], ["speech", 20.6], ["eruption", 11.35]], null, null, null, [["music", 67.71], ["hum", 4.34], ["electronic music", 3.81]]], "duration": [4.13, 0.48, 5.01, 33.55, 5.88, 1.89, 7.48]}

annotations_filtered/q4G5hUvL-wI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[21.0, 32.09], [33.0, 39.65], [42.0, 48.54], [49.0, 49.4], [50.0, 72.47], [75.0, 87.45], [88.0, 96.13], [96.0, 113.88]], "keep_status": [false, false, false, false, false, true, false, true], "silence_prob": [33.76, 32.35, 37.9, 0.0, 36.35, 34.87, 34.6, 34.0], "audiomae_on_audioset": [[["music", 64.56], ["synthesizer", 10.34], ["musical instrument", 4.38]], [["music", 59.36], ["musical instrument", 12.59], ["synthesizer", 5.86]], [["music", 68.37], ["effects unit", 9.12], ["guitar", 5.57]], null, [["music", 63.74], ["synthesizer", 4.78], ["musical instrument", 2.77]], [["music", 49.96], ["clarinet", 12.2], ["saxophone", 7.81]], [["music", 58.76], ["saxophone", 9.59], ["musical instrument", 4.51]], [["music", 55.42], ["saxophone", 7.38], ["harmonica", 7.17]]], "duration": [11.09, 6.65, 6.54, 0.4, 22.47, 12.45, 8.13, 17.88]}

annotations_filtered/q4Qlk7sfZfQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[7.0, 9.39], [11.0, 14.98], [15.0, 25.84], [27.0, 32.75], [36.0, 43.77], [49.0, 76.6], [79.0, 82.76], [84.0, 83.91], [89.0, 89.56], [90.0, 89.82], [90.0, 90.95], [91.0, 92.16], [92.0, 95.84], [96.0, 96.82], [101.0, 117.51], [119.0, 123.97], [127.0, 127.36]], "keep_status": [true, true, true, true, true, true, true, false, false, false, false, false, false, false, true, true, false], "silence_prob": [39.25, 44.37, 31.19, 31.03, 30.41, 30.86, 43.13, 0.0, 0.0, 0.0, 0.0, 0.0, 52.05, 0.0, 30.36, 30.33, 0.0], "audiomae_on_audioset": [[["music", 16.67], ["effects unit", 9.06], ["musical instrument", 6.95]], [["effects unit", 22.56], ["distortion", 13.1], ["noise", 6.49]], [["speech", 9.02], ["echo", 7.98], ["boing", 7.29]], [["noise", 12.4], ["hum", 11.83], ["buzz", 11.35]], [["car", 15.08], ["vehicle", 14.31], ["crowd", 9.19]], [["animal", 18.12], ["fly, housefly", 14.96], ["insect", 13.04]], [["fly, housefly", 24.15], ["noise", 17.64], ["insect", 11.4]], null, null, null, null, null, null, null, [["car", 12.76], ["echo", 10.57], ["vehicle", 10.23]], [["speech", 32.39], ["cacophony", 16.01], ["radio", 6.05]], null], "duration": [2.39, 3.98, 10.84, 5.75, 7.77, 27.6, 3.76, -0.09, 0.56, -0.18, 0.95, 1.16, 3.84, 0.82, 16.51, 4.97, 0.36]}

annotations_filtered/q4RbzjuXB6E_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 31.16], [33.0, 37.44], [40.0, 45.69], [46.0, 69.31], [82.0, 82.95], [84.0, 85.94], [88.0, 113.26], [114.0, 115.16]], "keep_status": [true, false, false, false, false, false, false, false], "silence_prob": [34.6, 47.2, 40.43, 35.36, 0.0, 0.0, 33.36, 0.0], "audiomae_on_audioset": [[["music", 31.26], ["speech", 20.23], ["electric shaver, electric razor", 9.69]], [["music", 48.4], ["didgeridoo", 15.03], ["speech", 10.6]], [["speech", 60.1], ["music", 23.52], ["fly, housefly", 2.28]], [["music", 64.08], ["throbbing", 5.84], ["hum", 3.34]], null, null, [["music", 48.63], ["speech", 26.18], ["electric shaver, electric razor", 3.78]], null], "duration": [26.16, 4.44, 5.69, 23.31, 0.95, 1.94, 25.26, 1.16]}

annotations_filtered/q4VIMzhfeYc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 9.09], [10.0, 16.53], [28.0, 29.56], [30.0, 38.58], [39.0, 49.77], [51.0, 51.41], [53.0, 54.23], [55.0, 56.56], [57.0, 65.74], [67.0, 68.12], [70.0, 70.66], [73.0, 72.89], [73.0, 73.06], [74.0, 74.9], [79.0, 82.53], [83.0, 82.95], [83.0, 83.17], [84.0, 84.2], [86.0, 86.93], [88.0, 88.89], [90.0, 98.1], [99.0, 103.47], [104.0, 105.56], [106.0, 107.03], [108.0, 108.94], [110.0, 111.42], [112.0, 113.04], [116.0, 116.61], [118.0, 118.5], [122.0, 123.42]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [96.54, 87.0, 0.0, 99.91, 100.0, 0.0, 0.0, 0.0, 78.55, 0.0, 0.0, 0.0, 0.0, 0.0, 85.72, 0.0, 0.0, 0.0, 0.0, 0.0, 99.99, 97.73, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [5.09, 6.53, 1.56, 8.58, 10.77, 0.41, 1.23, 1.56, 8.74, 1.12, 0.66, -0.11, 0.06, 0.9, 3.53, -0.05, 0.17, 0.2, 0.93, 0.89, 8.1, 4.47, 1.56, 1.03, 0.94, 1.42, 1.04, 0.61, 0.5, 1.42]}

annotations_filtered/q5BzDVDotzI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 3.3], [4.0, 5.05], [5.0, 6.39], [7.0, 7.7], [10.0, 10.47], [13.0, 13.27], [14.0, 15.35], [20.0, 21.36], [23.0, 23.62], [26.0, 27.43], [28.0, 29.17], [30.0, 31.13], [32.0, 33.4], [35.0, 35.01], [36.0, 36.54], [38.0, 38.58], [39.0, 40.26], [44.0, 44.1], [47.0, 47.73], [48.0, 49.25], [51.0, 51.66], [56.0, 55.93]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.3, 1.05, 1.39, 0.7, 0.47, 0.27, 1.35, 1.36, 0.62, 1.43, 1.17, 1.13, 1.4, 0.01, 0.54, 0.58, 1.26, 0.1, 0.73, 1.25, 0.66, -0.07]}

annotations_filtered/q5K1fm56gI8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 2.29], [6.0, 6.66], [8.0, 10.29], [12.0, 12.07], [13.0, 17.36], [22.0, 23.11], [29.0, 30.11], [35.0, 36.12], [39.0, 39.85], [45.0, 45.76], [48.0, 61.45], [62.0, 62.85], [65.0, 66.53], [68.0, 70.16], [73.0, 117.76], [119.0, 119.84], [123.0, 171.17]], "keep_status": [false, false, false, false, false, false, false, false, false, false, true, false, false, true, false, false, false], "silence_prob": [0.0, 0.0, 45.72, 0.0, 37.31, 0.0, 0.0, 0.0, 0.0, 0.0, 30.54, 0.0, 0.0, 30.08, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, [["sidetone", 69.18], ["speech", 18.74], ["chirp tone", 2.29]], null, [["speech", 56.16], ["sidetone", 12.94], ["hum", 4.17]], null, null, null, null, null, [["music", 53.88], ["synthesizer", 8.64], ["effects unit", 4.6]], null, null, [["music", 23.68], ["hum", 15.37], ["noise", 11.27]], null, null, null], "duration": [0.29, 0.66, 2.29, 0.07, 4.36, 1.11, 1.11, 1.12, 0.85, 0.76, 13.45, 0.85, 1.53, 2.16, 44.76, 0.84, 48.17]}

annotations_filtered/q5RSKejDWo8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 2.24], [5.0, 5.07], [7.0, 9.42], [11.0, 17.31], [26.0, 27.53], [31.0, 31.41], [32.0, 32.49], [33.0, 33.79], [41.0, 40.86], [42.0, 42.82], [44.0, 44.1], [47.0, 48.25], [53.0, 54.24], [55.0, 56.94], [58.0, 60.0], [62.0, 65.6], [69.0, 69.01], [76.0, 76.4], [77.0, 77.06], [80.0, 79.95], [81.0, 81.48], [82.0, 83.07], [83.0, 83.59], [84.0, 87.07], [88.0, 88.64], [89.0, 90.17]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 85.72, 60.7, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.88, 99.26, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 85.72, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.24, 0.07, 2.42, 6.31, 1.53, 0.41, 0.49, 0.79, -0.14, 0.82, 0.1, 1.25, 1.24, 1.94, 2.0, 3.6, 0.01, 0.4, 0.06, -0.05, 0.48, 1.07, 0.59, 3.07, 0.64, 1.17]}

annotations_filtered/q5eGg_CgBPk_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 6.0], [14.0, 15.08], [20.0, 36.95], [59.0, 61.13], [65.0, 79.03], [79.0, 79.07], [79.0, 79.1], [79.0, 79.41], [79.0, 79.44], [85.0, 85.11], [92.0, 98.17], [100.0, 102.29], [113.0, 120.5], [128.0, 130.54], [133.0, 148.44]], "keep_status": [false, false, true, false, false, false, false, false, false, false, false, true, false, true, false], "silence_prob": [0.0, 0.0, 32.51, 32.22, 32.7, 0.0, 0.0, 0.0, 0.0, 0.0, 32.88, 34.17, 33.75, 31.01, 35.3], "audiomae_on_audioset": [null, null, [["music", 23.97], ["speech", 18.5], ["breaking", 15.7]], [["speech", 41.2], ["music", 32.48], ["boing", 3.31]], [["speech", 48.29], ["music", 35.28], ["explosion", 2.42]], null, null, null, null, null, [["music", 45.03], ["didgeridoo", 14.62], ["speech", 11.91]], [["music", 41.47], ["theremin", 4.79], ["musical instrument", 4.75]], [["speech", 45.67], ["music", 35.79], ["sidetone", 6.72]], [["music", 41.66], ["trombone", 8.49], ["speech", 7.33]], [["whale vocalization", 40.86], ["theremin", 20.63], ["music", 16.89]]], "duration": [0.0, 1.08, 16.95, 2.13, 14.03, 0.07, 0.1, 0.41, 0.44, 0.11, 6.17, 2.29, 7.5, 2.54, 15.44]}

annotations_filtered/q5v5DOEF45E_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 5.51], [8.0, 9.15], [12.0, 12.97], [14.0, 15.4], [20.0, 20.73], [25.0, 26.5], [41.0, 44.22], [54.0, 55.16], [59.0, 61.92], [66.0, 66.8], [68.0, 69.21], [70.0, 77.53], [78.0, 82.97], [84.0, 89.45], [92.0, 94.95], [110.0, 111.3], [112.0, 118.88], [120.0, 121.15]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 100.0, 0.0, 0.0, 44.49, 63.74, 53.84, 54.97, 0.0, 40.41, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, [["speech", 39.46], ["music", 25.58], ["whale vocalization", 7.14]], null, null, null, null, [["livestock, farm animals, working animals", 17.36], ["hum", 16.68], ["speech", 14.08]], null], "duration": [1.51, 1.15, 0.97, 1.4, 0.73, 1.5, 3.22, 1.16, 2.92, 0.8, 1.21, 7.53, 4.97, 5.45, 2.95, 1.3, 6.88, 1.15]}

annotations_filtered/q6ObhNBURyY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[29.0, 74.87], [76.0, 143.33], [144.0, 143.95], [144.0, 166.08]], "keep_status": [false, false, false, true], "silence_prob": [0.0, 0.0, 0.0, 28.5], "audiomae_on_audioset": [null, null, null, [["music", 19.68], ["vehicle", 13.41], ["car", 11.03]]], "duration": [45.87, 67.33, -0.05, 22.08]}

annotations_filtered/q6XF66xysgQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 6.17], [7.0, 9.39], [11.0, 14.76], [16.0, 19.97], [22.0, 26.91], [27.0, 32.1], [39.0, 40.93], [42.0, 44.79], [46.0, 48.27], [52.0, 53.7], [57.0, 59.0], [60.0, 65.53], [67.0, 68.72], [76.0, 79.54], [81.0, 82.41]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [87.37, 93.29, 99.76, 98.66, 99.26, 97.33, 0.0, 97.92, 99.84, 0.0, 99.84, 99.71, 0.0, 99.76, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [2.17, 2.39, 3.76, 3.97, 4.91, 5.1, 1.93, 2.79, 2.27, 1.7, 2.0, 5.53, 1.72, 3.54, 1.41]}

annotations_filtered/q6j_0vS_NNM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 4.58], [16.0, 18.57], [21.0, 22.17], [23.0, 27.63], [30.0, 36.73], [40.0, 41.08], [45.0, 49.71], [54.0, 54.53], [56.0, 58.04], [60.0, 64.32], [66.0, 76.3], [78.0, 81.97], [85.0, 85.72], [87.0, 99.3], [102.0, 109.26], [114.0, 113.86], [114.0, 113.9], [114.0, 117.98], [126.0, 127.23], [135.0, 136.34], [138.0, 144.41], [146.0, 152.27], [153.0, 154.09], [158.0, 161.25], [162.0, 163.02], [165.0, 173.42], [177.0, 177.04], [180.0, 182.07]], "keep_status": [false, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 47.27, 0.0, 47.12, 39.63, 0.0, 36.55, 0.0, 35.52, 38.93, 36.99, 39.83, 0.0, 35.77, 38.97, 0.0, 0.0, 41.89, 0.0, 0.0, 36.27, 35.59, 0.0, 34.88, 0.0, 37.05, 0.0, 46.64], "audiomae_on_audioset": [null, [["music", 56.9], ["fly, housefly", 5.47], ["insect", 2.84]], null, [["music", 63.09], ["bass guitar", 4.08], ["double bass", 2.84]], [["music", 74.2], ["bass guitar", 2.43], ["double bass", 2.09]], null, [["music", 77.85], ["musical instrument", 1.82], ["mantra", 1.59]], null, [["music", 73.24], ["humming", 3.63], ["carnatic music", 2.71]], [["music", 83.24], ["sad music", 1.23], ["musical instrument", 1.17]], [["music", 67.54], ["double bass", 7.28], ["cello", 5.9]], [["music", 69.54], ["mantra", 3.72], ["carnatic music", 2.64]], null, [["music", 79.48], ["sad music", 3.12], ["cello", 1.98]], [["music", 57.86], ["double bass", 11.95], ["cello", 9.96]], null, null, [["music", 62.64], ["theremin", 6.0], ["cello", 4.46]], null, null, [["music", 78.74], ["guitar", 3.5], ["plucked string instrument", 2.71]], [["music", 81.57], ["mantra", 1.87], ["musical instrument", 1.33]], null, [["music", 73.93], ["mantra", 2.11], ["vocal music", 1.46]], null, [["music", 79.52], ["double bass", 3.29], ["cello", 1.47]], null, [["music", 73.25], ["musical instrument", 1.8], ["scary music", 1.51]]], "duration": [0.58, 2.57, 1.17, 4.63, 6.73, 1.08, 4.71, 0.53, 2.04, 4.32, 10.3, 3.97, 0.72, 12.3, 7.26, -0.14, -0.1, 3.98, 1.23, 1.34, 6.41, 6.27, 1.09, 3.25, 1.02, 8.42, 0.04, 2.07]}

annotations_filtered/q6zi7XGjQQw_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[23.0, 24.51], [25.0, 25.84], [28.0, 28.26], [30.0, 30.13], [31.0, 32.83], [35.0, 35.7], [36.0, 36.59], [47.0, 47.78], [51.0, 51.06], [52.0, 52.34], [53.0, 53.47], [55.0, 55.43], [56.0, 56.69], [61.0, 61.33], [70.0, 70.87], [79.0, 78.75], [80.0, 80.77], [86.0, 87.51], [88.0, 88.99], [90.0, 90.63], [91.0, 92.03], [95.0, 95.86], [108.0, 108.4]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [1.51, 0.84, 0.26, 0.13, 1.83, 0.7, 0.59, 0.78, 0.06, 0.34, 0.47, 0.43, 0.69, 0.33, 0.87, -0.25, 0.77, 1.51, 0.99, 0.63, 1.03, 0.86, 0.4]}

annotations_filtered/q74RKOmIjC8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[0.0, 1.49], [3.0, 3.47], [5.0, 6.4], [7.0, 7.86], [14.0, 14.32], [16.0, 16.68], [21.0, 22.57], [24.0, 25.61], [26.0, 39.18], [43.0, 43.28], [46.0, 48.36], [50.0, 50.92], [54.0, 55.38], [57.0, 59.07], [61.0, 62.01], [65.0, 65.8], [68.0, 68.96], [69.0, 70.39], [73.0, 73.68], [75.0, 75.91], [76.0, 77.3], [79.0, 79.81], [80.0, 80.5], [83.0, 83.39], [84.0, 87.05], [87.0, 89.58], [92.0, 92.72], [95.0, 95.98], [99.0, 99.13], [100.0, 100.82], [103.0, 104.26], [107.0, 110.27], [113.0, 113.8]], "keep_status": [false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 33.45, 0.0, 99.31, 0.0, 0.0, 84.8, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.36, 43.13, 0.0, 0.0, 0.0, 0.0, 0.0, 83.16, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, [["fart", 23.88], ["fly, housefly", 14.05], ["speech", 10.05]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["whale vocalization", 42.51], ["speech", 6.36], ["dog", 5.23]], null, null, null, null, null, null, null], "duration": [1.49, 0.47, 1.4, 0.86, 0.32, 0.68, 1.57, 1.61, 13.18, 0.28, 2.36, 0.92, 1.38, 2.07, 1.01, 0.8, 0.96, 1.39, 0.68, 0.91, 1.3, 0.81, 0.5, 0.39, 3.05, 2.58, 0.72, 0.98, 0.13, 0.82, 1.26, 3.27, 0.8]}

annotations_filtered/q7CX_5D6y6E_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[47.0, 52.2]], "keep_status": [false], "silence_prob": [34.25], "audiomae_on_audioset": [[["music", 52.01], ["flute", 17.82], ["theremin", 14.7]]], "duration": [5.2]}

annotations_filtered/q7DHkw_5Wzw_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 7.08], [9.0, 15.35], [15.0, 15.4], [15.0, 15.45], [15.0, 15.99], [16.0, 18.77], [19.0, 19.87], [23.0, 23.06], [24.0, 25.96], [28.0, 28.73], [30.0, 30.67], [31.0, 32.39], [33.0, 33.37], [36.0, 41.96], [45.0, 46.13], [47.0, 50.14], [53.0, 53.16], [56.0, 57.03], [59.0, 62.29], [63.0, 64.44], [66.0, 67.0], [69.0, 75.78], [77.0, 77.16], [81.0, 81.87], [83.0, 85.21], [88.0, 89.72], [93.0, 94.56], [103.0, 103.65], [112.0, 111.89], [114.0, 114.47], [118.0, 118.32], [122.0, 125.61], [126.0, 127.82], [130.0, 132.17], [134.0, 137.69], [138.0, 140.43], [145.0, 146.85], [149.0, 149.98], [152.0, 152.41], [154.0, 154.36], [155.0, 156.27], [160.0, 168.32], [170.0, 170.8], [172.0, 178.44], [179.0, 179.24], [181.0, 181.01], [182.0, 187.91], [189.0, 190.21], [192.0, 193.82], [198.0, 198.61], [202.0, 204.31], [206.0, 208.47], [209.0, 209.75], [211.0, 212.04], [213.0, 214.64], [216.0, 242.4]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [84.25, 59.07, 0.0, 0.0, 0.0, 73.97, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 89.72, 0.0, 100.0, 0.0, 0.0, 99.68, 0.0, 0.0, 99.05, 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 100.0, 100.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.98, 0.0, 100.0, 0.0, 0.0, 99.73, 0.0, 0.0, 0.0, 98.99, 99.59, 0.0, 0.0, 0.0, 99.98], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [4.08, 6.35, 0.4, 0.45, 0.99, 2.77, 0.87, 0.06, 1.96, 0.73, 0.67, 1.39, 0.37, 5.96, 1.13, 3.14, 0.16, 1.03, 3.29, 1.44, 1.0, 6.78, 0.16, 0.87, 2.21, 1.72, 1.56, 0.65, -0.11, 0.47, 0.32, 3.61, 1.82, 2.17, 3.69, 2.43, 1.85, 0.98, 0.41, 0.36, 1.27, 8.32, 0.8, 6.44, 0.24, 0.01, 5.91, 1.21, 1.82, 0.61, 2.31, 2.47, 0.75, 1.04, 1.64, 26.4]}

annotations_filtered/q7QxVddVEW0_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 3.38], [6.0, 6.2], [8.0, 10.18], [11.0, 11.89], [15.0, 16.01], [17.0, 18.55], [21.0, 21.61], [25.0, 25.0], [28.0, 30.3], [34.0, 35.58], [40.0, 40.47], [50.0, 64.32], [65.0, 82.36], [84.0, 88.69], [89.0, 94.71], [97.0, 96.8], [97.0, 109.85], [110.0, 112.99], [114.0, 114.4], [116.0, 117.64], [119.0, 119.45], [121.0, 121.66], [123.0, 123.26], [125.0, 127.6], [130.0, 132.19]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 89.72, 0.0, 0.0, 0.0, 0.0, 0.0, 98.8, 0.0, 0.0, 34.23, 44.2, 99.97, 99.94, 0.0, 79.41, 63.96, 0.0, 0.0, 0.0, 0.0, 0.0, 98.86, 99.85], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, [["music", 46.94], ["smash, crash", 20.03], ["speech", 16.66]], [["music", 45.63], ["speech", 28.74], ["hum", 5.21]], null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.38, 0.2, 2.18, 0.89, 1.01, 1.55, 0.61, 0.0, 2.3, 1.58, 0.47, 14.32, 17.36, 4.69, 5.71, -0.2, 12.85, 2.99, 0.4, 1.64, 0.45, 0.66, 0.26, 2.6, 2.19]}

annotations_filtered/q7S2ckr4IkM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[27.0, 27.23], [29.0, 29.34], [32.0, 35.02], [35.0, 35.06], [35.0, 35.61], [37.0, 38.16], [45.0, 45.23], [48.0, 48.32], [53.0, 53.32], [54.0, 56.4], [62.0, 62.43], [69.0, 70.39], [71.0, 72.01], [75.0, 75.88], [77.0, 78.12], [80.0, 80.13], [82.0, 82.68], [86.0, 86.56], [89.0, 90.59]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 53.84, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 58.55, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.23, 0.34, 3.02, 0.06, 0.61, 1.16, 0.23, 0.32, 0.32, 2.4, 0.43, 1.39, 1.01, 0.88, 1.12, 0.13, 0.68, 0.56, 1.59]}

annotations_filtered/q7V1sM0VNaw_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[10.0, 14.25], [20.0, 26.97], [28.0, 28.56], [32.0, 33.72], [35.0, 35.51], [37.0, 38.28], [39.0, 39.92], [47.0, 48.24], [49.0, 49.98], [51.0, 52.05], [54.0, 54.5], [56.0, 56.47], [58.0, 59.64], [61.0, 63.48], [64.0, 64.84], [66.0, 67.56], [71.0, 71.96], [76.0, 77.26], [84.0, 84.47], [86.0, 86.49], [89.0, 92.13], [93.0, 93.33], [101.0, 101.97], [105.0, 105.51], [106.0, 106.54], [108.0, 111.92], [113.0, 113.78], [116.0, 115.6], [117.0, 118.64], [120.0, 124.77]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [35.86, 31.92, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 60.42, 0.0, 0.0, 0.0, 0.0, 58.98, 0.0, 0.0, 0.0, 29.84], "audiomae_on_audioset": [[["music", 54.8], ["sidetone", 11.94], ["speech", 9.01]], [["music", 78.58], ["throbbing", 3.97], ["musical instrument", 2.13]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 68.68], ["speech", 12.16], ["applause", 4.75]]], "duration": [4.25, 6.97, 0.56, 1.72, 0.51, 1.28, 0.92, 1.24, 0.98, 1.05, 0.5, 0.47, 1.64, 2.48, 0.84, 1.56, 0.96, 1.26, 0.47, 0.49, 3.13, 0.33, 0.97, 0.51, 0.54, 3.92, 0.78, -0.4, 1.64, 4.77]}

annotations_filtered/q7heVIEyvQ4_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[41.0, 44.25], [47.0, 48.27]], "keep_status": [true, false], "silence_prob": [34.88, 0.0], "audiomae_on_audioset": [[["music", 52.21], ["didgeridoo", 4.22], ["throbbing", 3.6]], null], "duration": [3.25, 1.27]}

annotations_filtered/q7qwqVbZSqE_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [], "keep_status": [], "silence_prob": [], "audiomae_on_audioset": [], "duration": []}

annotations_filtered/q7tLJC4pC14_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 6.34], [13.0, 14.03], [16.0, 23.46], [24.0, 31.4], [33.0, 33.76], [34.0, 35.18], [36.0, 36.34], [37.0, 40.46], [42.0, 43.12], [45.0, 46.43], [49.0, 49.59], [52.0, 53.28], [54.0, 55.0], [56.0, 59.1], [60.0, 61.3], [64.0, 65.7], [68.0, 68.39], [85.0, 85.72], [86.0, 87.0], [89.0, 93.73], [95.0, 95.62], [96.0, 96.57], [98.0, 107.25]], "keep_status": [false, false, true, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, true, false, false, true], "silence_prob": [0.0, 0.0, 32.29, 78.38, 0.0, 0.0, 0.0, 52.51, 0.0, 0.0, 0.0, 0.0, 0.0, 43.51, 0.0, 0.0, 0.0, 0.0, 0.0, 49.54, 0.0, 0.0, 31.69], "audiomae_on_audioset": [null, null, [["music", 54.84], ["electronic music", 4.13], ["speech", 2.56]], null, null, null, null, null, null, null, null, null, null, [["music", 31.79], ["speech", 15.51], ["fart", 9.13]], null, null, null, null, null, [["speech", 23.04], ["hum", 16.89], ["mains hum", 8.45]], null, null, [["speech", 16.6], ["animal", 12.71], ["beatboxing", 6.29]]], "duration": [0.34, 1.03, 7.46, 7.4, 0.76, 1.18, 0.34, 3.46, 1.12, 1.43, 0.59, 1.28, 1.0, 3.1, 1.3, 1.7, 0.39, 0.72, 1.0, 4.73, 0.62, 0.57, 9.25]}