WIP

annotations_filtered/g-2hB5Kd5aM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 11.01], [15.0, 24.98], [29.0, 32.04], [34.0, 35.07], [35.0, 36.36], [39.0, 46.18], [48.0, 52.66], [54.0, 54.68], [56.0, 64.3], [71.0, 107.16], [110.0, 110.51], [112.0, 116.24]], "keep_status": [true, false, false, false, false, true, true, false, false, false, false, false], "silence_prob": [30.36, 32.15, 34.63, 0.0, 0.0, 31.23, 30.74, 0.0, 30.0, 0.0, 0.0, 94.52], "audiomae_on_audioset": [[["music", 17.07], ["synthesizer", 15.61], ["distortion", 13.93]], [["sidetone", 42.31], ["speech", 29.62], ["hum", 6.64]], [["music", 88.73], ["electronic music", 1.39], ["techno", 1.35]], null, null, [["synthesizer", 33.37], ["music", 12.68], ["noise", 10.48]], [["music", 46.83], ["noise", 13.69], ["synthesizer", 9.47]], null, [["speech", 64.72], ["music", 11.57], ["sidetone", 4.79]], null, null, null], "duration": [5.01, 9.98, 3.04, 1.07, 1.36, 7.18, 4.66, 0.68, 8.3, 36.16, 0.51, 4.24]}

annotations_filtered/g-GJDgd7D8k_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[7.0, 15.62], [17.0, 20.14], [23.0, 24.68], [27.0, 27.51], [30.0, 31.53], [33.0, 34.1], [37.0, 57.64], [58.0, 69.25], [70.0, 101.07], [105.0, 105.85], [109.0, 108.84], [115.0, 121.54], [126.0, 127.58]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [100.0, 99.99, 0.0, 0.0, 0.0, 0.0, 100.0, 100.0, 0.0, 0.0, 0.0, 100.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [8.62, 3.14, 1.68, 0.51, 1.53, 1.1, 20.64, 11.25, 31.07, 0.85, -0.16, 6.54, 1.58]}

annotations_filtered/g-P53rME1xE_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[25.0, 73.94], [78.0, 122.93]], "keep_status": [false, false], "silence_prob": [0.0, 0.0], "audiomae_on_audioset": [null, null], "duration": [48.94, 44.93]}

annotations_filtered/g-Yufp_dafk_filtered.json

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

annotations_filtered/g-g4vCbZsDM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 4.48], [7.0, 7.1], [10.0, 10.84], [12.0, 15.72], [16.0, 27.84], [28.0, 30.38], [42.0, 44.09], [45.0, 46.26], [48.0, 48.0], [48.0, 48.78], [50.0, 49.69], [51.0, 104.33], [105.0, 115.5], [117.0, 121.95]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 99.97, 90.43, 58.98, 99.97, 0.0, 0.0, 0.0, 0.0, 0.0, 52.33, 100.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [1.48, 0.1, 0.84, 3.72, 11.84, 2.38, 2.09, 1.26, 0.0, 0.78, -0.31, 53.33, 10.5, 4.95]}

annotations_filtered/g-gebDSBFkY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[34.0, 33.88], [36.0, 36.05], [40.0, 48.32], [49.0, 73.94], [75.0, 75.83], [77.0, 81.55], [84.0, 88.92]], "keep_status": [false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 30.05, 30.64, 0.0, 57.09, 60.98], "audiomae_on_audioset": [null, null, [["speech", 44.19], ["music", 19.6], ["shatter", 10.21]], [["speech", 64.12], ["music", 7.08], ["sidetone", 5.98]], null, null, null], "duration": [-0.12, 0.05, 8.32, 24.94, 0.83, 4.55, 4.92]}

annotations_filtered/g-uc5_QEmuM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 4.55], [7.0, 7.69], [10.0, 10.52], [17.0, 19.7], [25.0, 25.73], [28.0, 28.27], [31.0, 33.66], [35.0, 36.04], [40.0, 40.36], [44.0, 68.99], [80.0, 79.98], [80.0, 85.19], [89.0, 92.82], [94.0, 95.34], [98.0, 102.41], [103.0, 108.16], [112.0, 119.85]], "keep_status": [false, false, false, true, false, false, true, false, false, false, false, false, true, false, false, true, false], "silence_prob": [0.0, 0.0, 0.0, 29.76, 0.0, 0.0, 35.94, 0.0, 0.0, 28.7, 0.0, 28.67, 28.79, 0.0, 29.34, 28.81, 28.83], "audiomae_on_audioset": [null, null, null, [["cattle, bovinae", 18.03], ["music", 16.66], ["moo", 16.44]], null, null, [["music", 57.61], ["synthesizer", 2.97], ["didgeridoo", 2.95]], null, null, [["music", 79.17], ["crowd", 4.78], ["electronic music", 1.58]], null, [["music", 67.11], ["didgeridoo", 7.88], ["rattle (instrument)", 2.77]], [["music", 59.65], ["didgeridoo", 7.01], ["motorcycle", 2.71]], null, [["music", 89.27], ["didgeridoo", 3.33], ["musical instrument", 0.92]], [["music", 52.33], ["throbbing", 6.19], ["didgeridoo", 5.6]], [["music", 85.62], ["ska", 1.64], ["musical instrument", 0.89]]], "duration": [0.55, 0.69, 0.52, 2.7, 0.73, 0.27, 2.66, 1.04, 0.36, 24.99, -0.02, 5.19, 3.82, 1.34, 4.41, 5.16, 7.85]}

annotations_filtered/g05Ja_89tOg_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 1.53], [4.0, 4.19], [5.0, 15.11], [15.0, 15.87], [19.0, 19.9], [23.0, 23.38], [25.0, 29.61], [35.0, 41.74], [44.0, 44.9], [47.0, 54.65], [56.0, 56.79], [66.0, 66.16], [67.0, 67.71], [68.0, 72.03], [74.0, 93.26], [94.0, 99.69], [102.0, 101.82], [104.0, 106.98], [108.0, 109.8], [111.0, 112.55]], "keep_status": [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, 44.55, 0.0, 0.0, 0.0, 61.77, 86.09, 0.0, 35.18, 0.0, 0.0, 0.0, 99.56, 74.29, 100.0, 0.0, 100.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, [["speech", 86.76], ["sidetone", 2.94], ["radio", 1.92]], null, null, null, null, null, null, [["speech", 87.84], ["sidetone", 3.44], ["stomach rumble", 0.71]], null, null, null, null, null, null, null, null, null, null], "duration": [0.53, 0.19, 10.11, 0.87, 0.9, 0.38, 4.61, 6.74, 0.9, 7.65, 0.79, 0.16, 0.71, 4.03, 19.26, 5.69, -0.18, 2.98, 1.8, 1.55]}

annotations_filtered/g0AMLVSBfSs_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[0.0, 0.45], [3.0, 3.3], [6.0, 5.85], [13.0, 13.46], [25.0, 25.17], [28.0, 28.36], [29.0, 29.71], [39.0, 39.48]], "keep_status": [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], "audiomae_on_audioset": [null, null, null, null, null, null, null, null], "duration": [0.45, 0.3, -0.15, 0.46, 0.17, 0.36, 0.71, 0.48]}

annotations_filtered/g0CFQF54ePo_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[0.0, 0.46], [8.0, 8.21], [9.0, 9.98], [24.0, 24.53], [25.0, 26.65], [27.0, 29.56], [33.0, 33.93], [35.0, 35.33], [38.0, 37.98], [52.0, 53.0], [53.0, 54.04], [81.0, 84.1], [110.0, 113.51], [114.0, 113.61], [114.0, 114.1], [116.0, 119.1]], "keep_status": [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, 78.04, 0.0, 0.0, 0.0, 0.0, 0.0, 34.46, 34.41, 0.0, 0.0, 58.38], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, [["cattle, bovinae", 30.92], ["moo", 27.83], ["speech", 13.68]], [["speech", 27.68], ["vehicle", 7.9], ["livestock, farm animals, working animals", 7.68]], null, null, null], "duration": [0.46, 0.21, 0.98, 0.53, 1.65, 2.56, 0.93, 0.33, -0.02, 1.0, 1.04, 3.1, 3.51, -0.39, 0.1, 3.1]}

annotations_filtered/g0HwVyKSC_8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 6.3], [10.0, 10.57], [13.0, 49.65], [51.0, 51.12], [54.0, 53.89], [62.0, 63.09], [66.0, 66.73], [69.0, 69.55], [71.0, 71.36], [72.0, 73.63], [75.0, 76.06], [84.0, 84.54], [91.0, 90.91], [96.0, 96.28], [102.0, 102.79], [108.0, 108.33], [110.0, 110.24], [112.0, 117.76]], "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, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 90.78], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [1.3, 0.57, 36.65, 0.12, -0.11, 1.09, 0.73, 0.55, 0.36, 1.63, 1.06, 0.54, -0.09, 0.28, 0.79, 0.33, 0.24, 5.76]}

annotations_filtered/g0TZztZJGRo_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[20.0, 29.37], [31.0, 31.48], [32.0, 35.61], [36.0, 67.85]], "keep_status": [true, false, true, false], "silence_prob": [33.59, 0.0, 28.31, 0.0], "audiomae_on_audioset": [[["music", 23.98], ["buzz", 17.16], ["electric shaver, electric razor", 14.06]], null, [["music", 55.46], ["speech", 6.77], ["burst, pop", 3.35]], null], "duration": [9.37, 0.48, 3.61, 31.85]}

annotations_filtered/g0UV6ug96c0_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 11.16], [17.0, 18.47], [21.0, 25.27], [26.0, 26.35], [31.0, 36.64], [56.0, 56.54], [65.0, 65.01], [67.0, 78.44], [83.0, 83.57], [99.0, 100.16]], "keep_status": [true, false, false, false, false, false, false, false, false, false], "silence_prob": [31.35, 0.0, 34.03, 0.0, 31.48, 0.0, 0.0, 29.7, 0.0, 0.0], "audiomae_on_audioset": [[["speech", 29.28], ["mains hum", 20.55], ["hum", 13.42]], null, [["speech", 43.76], ["sidetone", 32.84], ["radio", 5.56]], null, [["speech", 68.53], ["vehicle", 9.25], ["car", 2.11]], null, null, [["speech", 74.1], ["explosion", 6.04], ["music", 5.66]], null, null], "duration": [3.16, 1.47, 4.27, 0.35, 5.64, 0.54, 0.01, 11.44, 0.57, 1.16]}

annotations_filtered/g0j2dVuhr6s_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 4.92], [6.0, 9.73], [12.0, 13.36], [16.0, 17.0], [20.0, 19.95], [26.0, 26.69], [28.0, 29.74], [30.0, 31.18], [32.0, 32.75], [34.0, 34.77], [35.0, 37.17], [40.0, 40.24], [42.0, 42.99], [45.0, 46.92], [49.0, 50.3], [53.0, 58.31], [60.0, 60.64], [62.0, 63.59], [66.0, 67.56], [68.0, 69.18], [71.0, 70.68], [73.0, 73.62], [80.0, 80.49], [85.0, 87.83], [92.0, 92.01], [95.0, 95.66], [107.0, 114.57], [120.0, 123.01], [124.0, 124.77], [126.0, 126.06]], "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, true, true, false, false], "silence_prob": [0.0, 29.28, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 37.92, 0.0, 0.0, 0.0, 0.0, 62.99, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 85.54, 0.0, 0.0, 38.56, 39.28, 0.0, 0.0], "audiomae_on_audioset": [null, [["busy signal", 39.56], ["sine wave", 15.52], ["sidetone", 14.66]], null, null, null, null, null, null, null, null, [["speech", 48.88], ["music", 19.45], ["sidetone", 18.04]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["hum", 9.28], ["music", 8.59], ["mains hum", 6.5]], [["firecracker", 18.58], ["noise", 17.84], ["fireworks", 6.37]], null, null], "duration": [0.92, 3.73, 1.36, 1.0, -0.05, 0.69, 1.74, 1.18, 0.75, 0.77, 2.17, 0.24, 0.99, 1.92, 1.3, 5.31, 0.64, 1.59, 1.56, 1.18, -0.32, 0.62, 0.49, 2.83, 0.01, 0.66, 7.57, 3.01, 0.77, 0.06]}

annotations_filtered/g0mHVE8ebqA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 3.45], [7.0, 7.03], [8.0, 10.98], [13.0, 13.26], [15.0, 16.75], [19.0, 20.73], [29.0, 29.89], [31.0, 32.05], [41.0, 43.14], [49.0, 48.96], [59.0, 61.77], [64.0, 65.65], [67.0, 67.69], [70.0, 75.41], [77.0, 77.6]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 99.68, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 100.0, 0.0, 0.0, 100.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.45, 0.03, 2.98, 0.26, 1.75, 1.73, 0.89, 1.05, 2.14, -0.04, 2.77, 1.65, 0.69, 5.41, 0.6]}

annotations_filtered/g0nhEzoCkJo_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[89.0, 93.16], [94.0, 114.67], [117.0, 117.75], [119.0, 122.08], [123.0, 129.0], [130.0, 131.28]], "keep_status": [true, true, false, true, false, false], "silence_prob": [38.51, 32.24, 0.0, 29.39, 55.89, 0.0], "audiomae_on_audioset": [[["whale vocalization", 13.65], ["livestock, farm animals, working animals", 11.27], ["music", 10.4]], [["music", 27.56], ["cattle, bovinae", 14.43], ["livestock, farm animals, working animals", 14.33]], null, [["mosquito", 22.21], ["insect", 18.74], ["fly, housefly", 16.96]], null, null], "duration": [4.16, 20.67, 0.75, 3.08, 6.0, 1.28]}

annotations_filtered/g0yYxO89lQA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[20.0, 23.33], [43.0, 43.34], [44.0, 44.39], [48.0, 84.99], [85.0, 86.29], [90.0, 90.9], [97.0, 98.25], [101.0, 100.99], [118.0, 118.08], [121.0, 122.22]], "keep_status": [false, false, false, false, false, false, false, false, false, false], "silence_prob": [65.32, 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], "duration": [3.33, 0.34, 0.39, 36.99, 1.29, 0.9, 1.25, -0.01, 0.08, 1.22]}

annotations_filtered/g1JAILio6-s_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 8.14], [51.0, 52.02], [54.0, 57.75], [66.0, 69.13], [72.0, 74.82], [76.0, 122.47]], "keep_status": [false, false, false, true, false, false], "silence_prob": [0.0, 0.0, 29.89, 36.25, 70.16, 0.0], "audiomae_on_audioset": [null, null, [["hum", 36.79], ["mains hum", 34.25], ["speech", 12.33]], [["whale vocalization", 39.87], ["hum", 8.25], ["music", 6.86]], null, null], "duration": [0.14, 1.02, 3.75, 3.13, 2.82, 46.47]}

annotations_filtered/g1jO4_HQQX4_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 6.35], [9.0, 12.9], [21.0, 21.29], [25.0, 25.76], [28.0, 28.32], [39.0, 39.63], [41.0, 41.89], [42.0, 43.6], [45.0, 45.47], [46.0, 62.29], [67.0, 67.24], [74.0, 87.1], [88.0, 93.07]], "keep_status": [false, true, false, false, false, false, false, false, false, true, false, true, true], "silence_prob": [0.0, 31.25, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 28.48, 0.0, 29.71, 28.67], "audiomae_on_audioset": [null, [["hum", 31.29], ["speech", 11.97], ["music", 9.27]], null, null, null, null, null, null, null, [["music", 46.74], ["hum", 10.28], ["throbbing", 9.32]], null, [["hum", 30.43], ["mains hum", 15.89], ["speech", 12.65]], [["fly, housefly", 33.14], ["mosquito", 11.65], ["whack, thwack", 10.69]]], "duration": [0.35, 3.9, 0.29, 0.76, 0.32, 0.63, 0.89, 1.6, 0.47, 16.29, 0.24, 13.1, 5.07]}

annotations_filtered/g1lpI9wZtiI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 15.36], [20.0, 28.56], [31.0, 48.39], [52.0, 53.32], [58.0, 58.08], [72.0, 84.74], [86.0, 173.25]], "keep_status": [false, false, false, false, false, true, false], "silence_prob": [29.85, 31.44, 64.75, 0.0, 0.0, 35.16, 0.0], "audiomae_on_audioset": [[["music", 43.07], ["hum", 25.29], ["mains hum", 13.43]], [["hum", 40.97], ["mains hum", 20.74], ["music", 11.19]], null, null, null, [["music", 44.82], ["theremin", 19.12], ["didgeridoo", 2.72]], null], "duration": [9.36, 8.56, 17.39, 1.32, 0.08, 12.74, 87.25]}

annotations_filtered/g1r-B5ZGZWY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[51.0, 51.44], [52.0, 52.73], [54.0, 54.23], [55.0, 56.29], [57.0, 59.0], [60.0, 66.29], [68.0, 89.51], [91.0, 119.03], [120.0, 138.47], [139.0, 140.8], [143.0, 143.02], [144.0, 145.1], [148.0, 148.53], [154.0, 154.5], [157.0, 164.22], [175.0, 188.94], [190.0, 196.55], [197.0, 198.36], [201.0, 201.75]], "keep_status": [false, false, false, false, false, true, true, false, false, false, false, false, false, false, true, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 87.74, 32.56, 32.42, 32.89, 80.11, 0.0, 0.0, 0.0, 0.0, 0.0, 38.98, 35.96, 93.13, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, [["hum", 25.45], ["music", 18.08], ["throbbing", 17.42]], [["music", 33.56], ["speech", 25.97], ["buzz", 4.36]], [["music", 31.59], ["throbbing", 30.15], ["hum", 17.59]], null, null, null, null, null, null, [["music", 57.22], ["speech", 9.47], ["frog", 2.85]], [["music", 64.38], ["speech", 12.64], ["bleat", 2.03]], null, null, null], "duration": [0.44, 0.73, 0.23, 1.29, 2.0, 6.29, 21.51, 28.03, 18.47, 1.8, 0.02, 1.1, 0.53, 0.5, 7.22, 13.94, 6.55, 1.36, 0.75]}

annotations_filtered/g2atr8aQ0zg_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 3.74], [6.0, 7.4], [10.0, 12.11], [17.0, 18.77], [19.0, 20.36], [22.0, 24.21], [30.0, 30.57], [33.0, 35.06], [36.0, 37.98], [39.0, 40.73], [42.0, 45.0], [46.0, 48.27], [53.0, 54.19], [55.0, 54.8], [63.0, 63.02], [67.0, 68.98], [72.0, 72.45], [73.0, 73.2], [76.0, 76.57], [83.0, 84.77], [86.0, 86.41], [88.0, 88.84], [91.0, 91.25], [94.0, 97.29], [98.0, 97.82], [101.0, 102.15], [103.0, 106.66], [110.0, 115.69], [117.0, 117.41], [123.0, 126.91], [127.0, 129.34]], "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], "silence_prob": [0.0, 0.0, 99.99, 0.0, 0.0, 94.37, 0.0, 84.25, 0.0, 0.0, 99.84, 93.76, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.48, 0.0, 0.0, 90.43, 99.26, 0.0, 85.17, 69.61], "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], "duration": [1.74, 1.4, 2.11, 1.77, 1.36, 2.21, 0.57, 2.06, 1.98, 1.73, 3.0, 2.27, 1.19, -0.2, 0.02, 1.98, 0.45, 0.2, 0.57, 1.77, 0.41, 0.84, 0.25, 3.29, -0.18, 1.15, 3.66, 5.69, 0.41, 3.91, 2.34]}

annotations_filtered/g2dAymk715E_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 2.91], [14.0, 14.94], [27.0, 33.02], [35.0, 36.56], [38.0, 39.06], [45.0, 45.79], [55.0, 55.56], [67.0, 73.04], [76.0, 80.87], [83.0, 83.03], [86.0, 87.1], [106.0, 106.79], [122.0, 122.88], [138.0, 139.4], [151.0, 151.56], [154.0, 158.14], [160.0, 161.23], [166.0, 165.87], [174.0, 174.38]], "keep_status": [false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 42.08, 0.0, 0.0, 0.0, 0.0, 34.13, 42.11, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.94, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, [["sidetone", 64.55], ["speech", 28.04], ["hum", 1.24]], null, null, null, null, [["speech", 48.5], ["whack, thwack", 22.07], ["thunk", 12.64]], [["speech", 19.26], ["hum", 7.64], ["fart", 7.14]], null, null, null, null, null, null, null, null, null, null], "duration": [-0.09, 0.94, 6.02, 1.56, 1.06, 0.79, 0.56, 6.04, 4.87, 0.03, 1.1, 0.79, 0.88, 1.4, 0.56, 4.14, 1.23, -0.13, 0.38]}

annotations_filtered/g2h8xRzMxtA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 4.87], [10.0, 10.5], [13.0, 14.23], [16.0, 16.17], [17.0, 17.47], [23.0, 24.46], [26.0, 34.69], [36.0, 39.09], [40.0, 41.35], [42.0, 42.35], [43.0, 45.3], [47.0, 48.0], [51.0, 51.46], [59.0, 60.96]], "keep_status": [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, 99.05, 98.19, 0.0, 0.0, 98.01, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.87, 0.5, 1.23, 0.17, 0.47, 1.46, 8.69, 3.09, 1.35, 0.35, 2.3, 1.0, 0.46, 1.96]}

annotations_filtered/g2iWVWVSb6Q_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[21.0, 24.49], [31.0, 31.36], [32.0, 32.49], [34.0, 34.43], [36.0, 37.15], [37.0, 38.87], [43.0, 43.56], [53.0, 54.06], [55.0, 55.58], [63.0, 64.4], [65.0, 64.94], [74.0, 73.99], [82.0, 83.24], [84.0, 83.96], [85.0, 85.33], [86.0, 97.53], [101.0, 101.75]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [93.29, 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, 99.96, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [3.49, 0.36, 0.49, 0.43, 1.15, 1.87, 0.56, 1.06, 0.58, 1.4, -0.06, -0.01, 1.24, -0.04, 0.33, 11.53, 0.75]}

annotations_filtered/g2tNQ_6-kpg_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 6.22], [7.0, 8.13], [10.0, 11.3], [13.0, 14.25], [15.0, 15.75], [17.0, 25.17], [26.0, 29.74], [30.0, 33.71], [36.0, 40.58], [42.0, 43.75], [44.0, 46.38], [47.0, 49.72], [51.0, 51.85], [55.0, 57.15], [59.0, 60.76], [62.0, 62.5], [63.0, 63.76], [65.0, 65.42], [67.0, 72.66], [75.0, 85.16], [86.0, 91.52], [92.0, 91.57], [92.0, 96.47], [98.0, 100.35], [101.0, 104.48], [107.0, 107.74], [109.0, 110.76], [113.0, 112.94], [114.0, 115.5], [117.0, 123.69]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, true], "silence_prob": [72.9, 0.0, 0.0, 0.0, 0.0, 65.67, 66.15, 59.15, 78.38, 0.0, 65.55, 53.65, 0.0, 82.43, 0.0, 0.0, 0.0, 0.0, 55.18, 53.1, 61.67, 0.0, 44.84, 100.0, 100.0, 0.0, 0.0, 0.0, 0.0, 41.89], "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, [["mains hum", 21.25], ["noise", 12.27], ["hum", 11.54]], null, null, null, null, null, null, [["noise", 35.53], ["speech", 11.33], ["music", 7.73]]], "duration": [2.22, 1.13, 1.3, 1.25, 0.75, 8.17, 3.74, 3.71, 4.58, 1.75, 2.38, 2.72, 0.85, 2.15, 1.76, 0.5, 0.76, 0.42, 5.66, 10.16, 5.52, -0.43, 4.47, 2.35, 3.48, 0.74, 1.76, -0.06, 1.5, 6.69]}

annotations_filtered/g3D2eGiLoeI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[11.0, 36.73], [39.0, 53.43], [55.0, 55.16], [56.0, 63.29], [67.0, 68.57], [76.0, 77.08], [78.0, 78.92], [85.0, 86.48], [88.0, 88.94], [90.0, 92.11], [93.0, 95.54], [98.0, 98.76], [99.0, 100.77], [101.0, 102.74], [104.0, 106.19]], "keep_status": [false, true, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [75.55, 33.94, 0.0, 99.62, 0.0, 0.0, 0.0, 0.0, 0.0, 99.91, 99.99, 0.0, 0.0, 0.0, 66.15], "audiomae_on_audioset": [null, [["speech", 51.88], ["fart", 5.09], ["aircraft", 3.92]], null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [25.73, 14.43, 0.16, 7.29, 1.57, 1.08, 0.92, 1.48, 0.94, 2.11, 2.54, 0.76, 1.77, 1.74, 2.19]}

annotations_filtered/g3E69dpurZA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[17.0, 17.07], [17.0, 17.44], [17.0, 21.95]], "keep_status": [false, false, false], "silence_prob": [0.0, 0.0, 32.71], "audiomae_on_audioset": [null, null, [["speech", 75.89], ["sidetone", 7.95], ["radio", 2.05]]], "duration": [0.07, 0.44, 4.95]}

annotations_filtered/g3FFfmWvyAk_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 7.48], [9.0, 10.45], [11.0, 11.99], [18.0, 19.45], [22.0, 24.16], [27.0, 27.18], [28.0, 30.7], [38.0, 39.5], [43.0, 50.48], [51.0, 53.27], [54.0, 61.7], [63.0, 63.41], [64.0, 64.49], [71.0, 71.64], [72.0, 74.31], [98.0, 99.47], [111.0, 111.05], [114.0, 114.64]], "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, 100.0, 0.0, 65.2, 0.0, 68.15, 99.94, 80.29, 0.0, 0.0, 0.0, 86.27, 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], "duration": [1.48, 1.45, 0.99, 1.45, 2.16, 0.18, 2.7, 1.5, 7.48, 2.27, 7.7, 0.41, 0.49, 0.64, 2.31, 1.47, 0.05, 0.64]}

annotations_filtered/g3WSsm57iVM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 13.56], [16.0, 21.1], [22.0, 23.46], [26.0, 27.6], [29.0, 34.37], [35.0, 39.04], [41.0, 40.85], [44.0, 44.56], [46.0, 48.02], [49.0, 51.33], [52.0, 65.3], [66.0, 70.12], [72.0, 72.35], [73.0, 79.49]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [94.95, 100.0, 0.0, 0.0, 94.07, 90.08, 0.0, 0.0, 99.94, 99.97, 99.96, 98.51, 0.0, 91.64], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [11.56, 5.1, 1.46, 1.6, 5.37, 4.04, -0.15, 0.56, 2.02, 2.33, 13.3, 4.12, 0.35, 6.49]}

annotations_filtered/g3WtvzmKCQQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 2.61], [4.0, 11.82], [15.0, 17.15], [18.0, 18.49], [20.0, 21.44], [23.0, 25.98], [28.0, 30.3], [32.0, 33.13], [36.0, 37.23], [39.0, 40.49], [42.0, 42.23], [45.0, 47.01], [47.0, 47.97], [49.0, 50.94], [53.0, 52.91], [54.0, 54.48], [55.0, 55.46], [56.0, 57.99], [59.0, 62.45], [63.0, 63.61], [65.0, 66.02], [68.0, 76.6], [77.0, 78.02], [82.0, 83.84], [86.0, 86.49], [91.0, 93.85], [95.0, 97.53], [99.0, 99.1], [102.0, 104.82]], "keep_status": [false, true, true, false, false, true, true, false, false, false, false, false, false, false, false, false, false, false, true, false, false, true, false, false, false, true, false, false, true], "silence_prob": [0.0, 42.13, 41.83, 0.0, 0.0, 38.42, 40.4, 0.0, 0.0, 0.0, 0.0, 40.57, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 30.34, 0.0, 0.0, 29.54, 0.0, 0.0, 0.0, 44.29, 42.37, 0.0, 37.91], "audiomae_on_audioset": [null, [["noise", 12.06], ["synthesizer", 11.57], ["hum", 10.32]], [["hum", 17.08], ["rumble", 11.86], ["mains hum", 6.18]], null, null, [["speech", 38.41], ["hum", 19.53], ["mains hum", 9.47]], [["music", 15.18], ["hum", 13.02], ["effects unit", 8.83]], null, null, null, null, [["hum", 55.08], ["mains hum", 17.07], ["throbbing", 9.93]], null, null, null, null, null, null, [["hum", 37.61], ["mains hum", 14.15], ["noise", 8.63]], null, null, [["music", 43.38], ["guitar", 10.65], ["musical instrument", 7.22]], null, null, null, [["hum", 32.47], ["throbbing", 25.86], ["whale vocalization", 11.31]], [["speech", 27.79], ["whale vocalization", 23.29], ["sidetone", 19.44]], null, [["throbbing", 33.47], ["hum", 24.13], ["music", 6.43]]], "duration": [0.61, 7.82, 2.15, 0.49, 1.44, 2.98, 2.3, 1.13, 1.23, 1.49, 0.23, 2.01, 0.97, 1.94, -0.09, 0.48, 0.46, 1.99, 3.45, 0.61, 1.02, 8.6, 1.02, 1.84, 0.49, 2.85, 2.53, 0.1, 2.82]}

annotations_filtered/g3a9qZnTzJQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 5.21], [7.0, 9.78], [13.0, 13.24], [20.0, 26.33], [28.0, 41.84], [48.0, 48.03], [51.0, 50.67], [52.0, 53.1], [59.0, 63.09], [65.0, 64.79], [67.0, 67.04], [69.0, 83.56], [84.0, 85.04], [86.0, 88.1], [89.0, 94.71], [95.0, 96.04], [97.0, 122.25], [123.0, 122.94], [124.0, 134.59]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false, true, false, true], "silence_prob": [0.0, 68.8, 0.0, 59.86, 59.33, 0.0, 0.0, 0.0, 78.55, 0.0, 0.0, 79.94, 0.0, 46.72, 33.45, 0.0, 34.25, 0.0, 34.56], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 59.14], ["ambient music", 11.19], ["electronic music", 6.99]], [["speech", 17.56], ["groan", 14.67], ["music", 13.32]], null, [["buzz", 17.09], ["fly, housefly", 12.75], ["speech", 11.45]], null, [["speech", 38.99], ["music", 19.22], ["thump, thud", 4.81]]], "duration": [1.21, 2.78, 0.24, 6.33, 13.84, 0.03, -0.33, 1.1, 4.09, -0.21, 0.04, 14.56, 1.04, 2.1, 5.71, 1.04, 25.25, -0.06, 10.59]}

annotations_filtered/g3hYbDHwBJY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[12.0, 21.17], [23.0, 25.24], [33.0, 39.26], [39.0, 47.17], [48.0, 49.52], [51.0, 51.12], [58.0, 58.19], [60.0, 61.25], [62.0, 62.46], [63.0, 63.49], [102.0, 109.02], [110.0, 110.3]], "keep_status": [false, true, true, false, false, false, false, false, false, false, true, false], "silence_prob": [30.62, 33.11, 30.32, 52.39, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 34.59, 0.0], "audiomae_on_audioset": [[["music", 51.76], ["speech", 19.34], ["fly, housefly", 7.84]], [["music", 41.48], ["speech", 10.56], ["hum", 10.39]], [["throbbing", 24.33], ["hum", 14.73], ["music", 8.76]], null, null, null, null, null, null, null, [["hum", 30.97], ["throbbing", 13.7], ["mains hum", 8.89]], null], "duration": [9.17, 2.24, 6.26, 8.17, 1.52, 0.12, 0.19, 1.25, 0.46, 0.49, 7.02, 0.3]}

annotations_filtered/g3jImb6V4wI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 6.57], [8.0, 11.26], [20.0, 20.41], [24.0, 24.95], [26.0, 28.81], [32.0, 33.03], [35.0, 34.89], [36.0, 36.54], [38.0, 38.11], [40.0, 41.37], [42.0, 42.96], [45.0, 47.83], [48.0, 48.59], [50.0, 51.54], [57.0, 60.76], [61.0, 62.21], [63.0, 63.29], [65.0, 65.99], [67.0, 67.66], [68.0, 67.91], [69.0, 69.67], [73.0, 74.61], [75.0, 76.18], [93.0, 95.18]], "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, true], "silence_prob": [0.0, 47.7, 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 62.99, 0.0, 0.0, 58.38, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 35.99], "audiomae_on_audioset": [null, [["speech", 73.29], ["synthesizer", 6.01], ["music", 3.61]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 22.04], ["bass guitar", 13.79], ["guitar", 12.38]]], "duration": [0.57, 3.26, 0.41, 0.95, 2.81, 1.03, -0.11, 0.54, 0.11, 1.37, 0.96, 2.83, 0.59, 1.54, 3.76, 1.21, 0.29, 0.99, 0.66, -0.09, 0.67, 1.61, 1.18, 2.18]}

annotations_filtered/g3kYdbqIwBE_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[12.0, 12.33], [17.0, 18.25], [21.0, 25.22], [35.0, 36.39], [49.0, 51.8], [55.0, 55.51], [64.0, 64.35], [68.0, 68.45], [88.0, 88.89], [90.0, 90.93]], "keep_status": [false, false, true, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 39.5, 0.0, 33.62, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, [["mosquito", 23.43], ["fly, housefly", 17.39], ["insect", 12.71]], null, [["boing", 73.83], ["speech", 8.73], ["cattle, bovinae", 2.73]], null, null, null, null, null], "duration": [0.33, 1.25, 4.22, 1.39, 2.8, 0.51, 0.35, 0.45, 0.89, 0.93]}

annotations_filtered/g3svdzmBtic_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[27.0, 39.75], [42.0, 61.38], [67.0, 71.66], [75.0, 115.2], [118.0, 125.42], [126.0, 127.5], [134.0, 138.25], [143.0, 142.67], [155.0, 159.68], [168.0, 173.3]], "keep_status": [true, true, true, false, false, false, true, false, true, true], "silence_prob": [28.88, 29.03, 29.95, 0.0, 30.08, 0.0, 29.42, 0.0, 29.37, 28.8], "audiomae_on_audioset": [[["speech", 37.51], ["fly, housefly", 13.03], ["music", 8.41]], [["music", 28.02], ["speech", 18.53], ["hum", 13.12]], [["speech", 46.15], ["music", 9.43], ["fart", 6.15]], null, [["music", 52.08], ["theremin", 22.94], ["scary music", 5.13]], null, [["hum", 30.47], ["mains hum", 21.69], ["music", 12.3]], null, [["mosquito", 23.02], ["speech", 19.58], ["fly, housefly", 12.74]], [["music", 46.9], ["speech", 4.63], ["musical instrument", 2.89]]], "duration": [12.75, 19.38, 4.66, 40.2, 7.42, 1.5, 4.25, -0.33, 4.68, 5.3]}

annotations_filtered/g425SDBoDBI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 7.57], [9.0, 9.48], [15.0, 15.58], [18.0, 20.6], [24.0, 25.19], [29.0, 29.22], [31.0, 31.68], [32.0, 32.88], [40.0, 42.28], [44.0, 45.5], [58.0, 58.35], [60.0, 60.0], [61.0, 61.53], [62.0, 63.26], [64.0, 64.52], [65.0, 65.5], [66.0, 67.34], [68.0, 69.55], [73.0, 76.06], [80.0, 80.32], [82.0, 86.95], [92.0, 92.58], [97.0, 96.77], [104.0, 104.45], [105.0, 105.88], [107.0, 201.11], [203.0, 205.34], [207.0, 208.26], [208.0, 211.67]], "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, true, false, true], "silence_prob": [59.59, 0.0, 0.0, 63.64, 0.0, 0.0, 0.0, 0.0, 75.55, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 75.07, 0.0, 40.62, 0.0, 0.0, 0.0, 0.0, 0.0, 39.63, 0.0, 31.36], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 48.9], ["throbbing", 13.75], ["hum", 9.29]], null, null, null, null, null, [["music", 43.05], ["ambient music", 12.78], ["hum", 9.34]], null, [["music", 40.59], ["ambient music", 14.58], ["gong", 7.39]]], "duration": [2.57, 0.48, 0.58, 2.6, 1.19, 0.22, 0.68, 0.88, 2.28, 1.5, 0.35, 0.0, 0.53, 1.26, 0.52, 0.5, 1.34, 1.55, 3.06, 0.32, 4.95, 0.58, -0.23, 0.45, 0.88, 94.11, 2.34, 1.26, 3.67]}

annotations_filtered/g46IxT3MGP8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 29.07], [30.0, 53.45], [54.0, 54.21], [54.0, 56.46], [57.0, 58.5], [59.0, 60.39], [62.0, 62.13], [63.0, 65.69], [67.0, 68.49], [72.0, 72.49], [80.0, 80.37], [81.0, 81.89], [83.0, 84.08], [88.0, 96.01], [97.0, 98.44], [102.0, 103.84], [105.0, 105.36], [106.0, 106.91], [109.0, 109.21], [111.0, 112.51], [114.0, 116.13], [117.0, 117.95], [119.0, 119.74]], "keep_status": [false, true, false, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, false, false, false], "silence_prob": [57.81, 35.73, 0.0, 52.92, 0.0, 0.0, 0.0, 59.59, 0.0, 0.0, 0.0, 0.0, 0.0, 41.36, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 80.46, 0.0, 0.0], "audiomae_on_audioset": [null, [["speech", 47.29], ["hum", 10.87], ["throbbing", 6.81]], null, null, null, null, null, null, null, null, null, null, null, [["speech", 28.58], ["music", 17.53], ["hum", 4.76]], null, null, null, null, null, null, null, null, null], "duration": [25.07, 23.45, 0.21, 2.46, 1.5, 1.39, 0.13, 2.69, 1.49, 0.49, 0.37, 0.89, 1.08, 8.01, 1.44, 1.84, 0.36, 0.91, 0.21, 1.51, 2.13, 0.95, 0.74]}

annotations_filtered/g4FOpeshqA8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 3.7], [6.0, 8.95], [11.0, 12.65], [15.0, 17.98], [19.0, 20.34], [22.0, 22.82], [36.0, 36.29], [49.0, 49.37], [51.0, 51.33], [56.0, 56.98], [58.0, 60.39], [65.0, 66.19], [72.0, 73.01], [82.0, 85.16], [89.0, 90.86], [94.0, 95.2], [97.0, 98.27], [104.0, 105.02], [111.0, 112.83], [116.0, 117.91], [120.0, 122.25], [124.0, 126.94], [129.0, 131.04]], "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, 91.13, 0.0, 68.41, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 36.42, 0.0, 0.0, 67.51, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 54.17, 52.22, 50.56], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, [["music", 73.36], ["synthesizer", 5.3], ["electronic music", 3.06]], null, null, null, null, null, null, null, null, null, null, null, null], "duration": [1.7, 2.95, 1.65, 2.98, 1.34, 0.82, 0.29, 0.37, 0.33, 0.98, 2.39, 1.19, 1.01, 3.16, 1.86, 1.2, 1.27, 1.02, 1.83, 1.91, 2.25, 2.94, 2.04]}

annotations_filtered/g4v51XeJnkw_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 1.9], [3.0, 6.08], [6.0, 8.83], [10.0, 22.84], [25.0, 30.33], [31.0, 35.16], [37.0, 40.95], [41.0, 41.35], [42.0, 47.44], [48.0, 48.95], [49.0, 54.43], [55.0, 57.25], [59.0, 60.78], [62.0, 63.39], [64.0, 64.69], [66.0, 85.29], [91.0, 91.17], [93.0, 94.09], [98.0, 99.0], [104.0, 105.04], [108.0, 110.69], [111.0, 111.67], [115.0, 116.43], [119.0, 119.77], [121.0, 123.55]], "keep_status": [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": [0.0, 49.4, 96.17, 39.57, 71.57, 30.88, 33.54, 0.0, 94.66, 0.0, 75.39, 53.16, 0.0, 0.0, 0.0, 42.67, 0.0, 0.0, 0.0, 0.0, 79.41, 0.0, 0.0, 0.0, 92.97], "audiomae_on_audioset": [null, [["music", 30.28], ["speech", 25.91], ["thunk", 8.08]], null, [["speech", 70.65], ["whack, thwack", 3.03], ["sidetone", 2.84]], null, [["fly, housefly", 40.98], ["mosquito", 13.44], ["speech", 6.86]], [["speech", 78.23], ["echo", 2.16], ["music", 1.63]], null, null, null, null, null, null, null, null, [["music", 65.04], ["speech", 8.52], ["throbbing", 4.89]], null, null, null, null, null, null, null, null, null], "duration": [0.9, 3.08, 2.83, 12.84, 5.33, 4.16, 3.95, 0.35, 5.44, 0.95, 5.43, 2.25, 1.78, 1.39, 0.69, 19.29, 0.17, 1.09, 1.0, 1.04, 2.69, 0.67, 1.43, 0.77, 2.55]}

annotations_filtered/g5-KsABvVzU_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 3.38], [4.0, 6.25], [7.0, 9.27], [9.0, 11.16], [13.0, 13.88], [15.0, 16.44], [18.0, 18.45], [20.0, 24.46], [26.0, 26.86], [28.0, 39.53], [40.0, 47.95], [51.0, 55.93], [58.0, 61.91], [67.0, 68.28], [70.0, 70.65], [79.0, 98.44], [100.0, 104.55], [106.0, 121.96], [123.0, 125.44], [128.0, 132.98], [134.0, 135.46], [139.0, 141.08], [146.0, 147.61], [151.0, 150.94], [152.0, 169.35], [171.0, 171.26], [171.0, 174.51]], "keep_status": [false, false, false, false, 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": [0.0, 100.0, 89.01, 98.51, 0.0, 0.0, 0.0, 45.72, 0.0, 50.31, 75.88, 96.42, 82.43, 0.0, 0.0, 53.34, 100.0, 62.07, 75.72, 67.51, 0.0, 50.41, 0.0, 0.0, 41.16, 0.0, 35.94], "audiomae_on_audioset": [null, null, null, null, null, null, null, [["stomach rumble", 50.44], ["whale vocalization", 7.97], ["burping, eructation", 5.1]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 54.8], ["throbbing", 12.29], ["hum", 5.3]], null, [["speech", 54.32], ["whale vocalization", 13.08], ["music", 6.62]]], "duration": [1.38, 2.25, 2.27, 2.16, 0.88, 1.44, 0.45, 4.46, 0.86, 11.53, 7.95, 4.93, 3.91, 1.28, 0.65, 19.44, 4.55, 15.96, 2.44, 4.98, 1.46, 2.08, 1.61, -0.06, 17.35, 0.26, 3.51]}

annotations_filtered/g511NYTRiOE_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 5.97], [6.0, 7.52], [8.0, 12.38], [13.0, 14.07], [15.0, 16.6], [17.0, 17.2], [17.0, 29.94], [32.0, 35.94], [37.0, 39.7], [40.0, 42.65], [45.0, 46.77], [50.0, 54.08], [59.0, 60.59], [61.0, 63.81], [65.0, 67.78], [68.0, 69.89], [83.0, 85.46], [90.0, 93.39], [95.0, 95.61], [98.0, 98.14], [99.0, 99.42], [101.0, 101.65], [104.0, 105.09], [114.0, 113.9], [122.0, 127.3], [128.0, 131.63], [132.0, 134.22], [135.0, 136.0], [138.0, 140.29], [146.0, 147.28], [151.0, 151.75], [154.0, 154.08], [169.0, 168.59], [169.0, 169.87], [190.0, 189.94], [192.0, 191.91], [196.0, 196.57], [198.0, 198.68], [201.0, 201.99], [203.0, 203.23]], "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, true, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [68.67, 0.0, 62.78, 0.0, 0.0, 0.0, 55.89, 97.92, 96.54, 59.07, 0.0, 98.8, 0.0, 98.66, 47.5, 0.0, 96.04, 74.92, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 99.48, 97.22, 39.54, 0.0, 35.2, 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, [["sidetone", 53.62], ["speech", 19.83], ["sine wave", 5.3]], null, null, null, null, null, null, null, null, null, null, null, [["speech", 57.49], ["hum", 11.99], ["throbbing", 5.89]], null, [["music", 28.54], ["musical instrument", 21.74], ["drum", 10.62]], null, null, null, null, null, null, null, null, null, null, null], "duration": [4.97, 1.52, 4.38, 1.07, 1.6, 0.2, 12.94, 3.94, 2.7, 2.65, 1.77, 4.08, 1.59, 2.81, 2.78, 1.89, 2.46, 3.39, 0.61, 0.14, 0.42, 0.65, 1.09, -0.1, 5.3, 3.63, 2.22, 1.0, 2.29, 1.28, 0.75, 0.08, -0.41, 0.87, -0.06, -0.09, 0.57, 0.68, 0.99, 0.23]}

annotations_filtered/g5AixBKy7b4_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 3.72], [6.0, 5.81], [14.0, 15.38], [21.0, 23.94], [26.0, 27.92], [30.0, 33.34], [35.0, 36.41], [38.0, 39.23], [40.0, 41.39], [44.0, 52.89], [53.0, 54.73], [57.0, 59.12]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 65.2, 0.0, 75.88, 0.0, 0.0, 0.0, 40.07, 0.0, 98.51], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, [["music", 63.66], ["speech", 12.24], ["synthesizer", 7.32]], null, null], "duration": [0.72, -0.19, 1.38, 2.94, 1.92, 3.34, 1.41, 1.23, 1.39, 8.89, 1.73, 2.12]}

annotations_filtered/g5Y-PN_duno_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 3.37], [16.0, 17.88], [23.0, 23.21], [24.0, 25.25], [26.0, 27.31], [29.0, 30.77], [32.0, 33.54], [36.0, 36.73], [39.0, 39.04], [43.0, 43.7], [45.0, 46.31], [48.0, 49.77], [52.0, 53.92], [55.0, 54.73], [56.0, 56.2], [61.0, 61.72], [64.0, 65.15], [69.0, 69.99], [74.0, 77.65], [80.0, 81.55], [82.0, 82.43], [86.0, 86.95], [89.0, 88.94], [92.0, 93.39], [95.0, 95.35], [98.0, 98.49], [103.0, 104.55], [108.0, 111.64], [113.0, 114.83]], "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], "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, 91.47, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 55.31, 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], "duration": [0.37, 1.88, 0.21, 1.25, 1.31, 1.77, 1.54, 0.73, 0.04, 0.7, 1.31, 1.77, 1.92, -0.27, 0.2, 0.72, 1.15, 0.99, 3.65, 1.55, 0.43, 0.95, -0.06, 1.39, 0.35, 0.49, 1.55, 3.64, 1.83]}

annotations_filtered/g5e3qoREpuA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 3.5], [5.0, 12.19], [13.0, 14.99], [16.0, 18.99], [19.0, 21.05], [22.0, 22.45], [23.0, 24.44], [26.0, 27.77], [32.0, 32.58], [34.0, 34.7], [42.0, 48.49], [50.0, 52.1], [55.0, 55.41], [56.0, 68.61], [72.0, 87.54], [90.0, 95.88], [97.0, 97.53], [98.0, 99.86], [108.0, 141.03], [146.0, 147.55], [150.0, 161.2], [162.0, 162.53], [163.0, 163.83], [165.0, 165.96], [167.0, 168.34], [168.0, 168.4], [168.0, 170.06]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, false, false, false, false, true, false, false, false, false, false, true], "silence_prob": [0.0, 56.86, 0.0, 68.8, 55.18, 0.0, 0.0, 0.0, 0.0, 0.0, 90.78, 57.48, 0.0, 33.01, 31.1, 30.8, 0.0, 0.0, 0.0, 0.0, 39.19, 0.0, 0.0, 0.0, 0.0, 0.0, 35.03], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, [["hum", 28.15], ["mains hum", 17.68], ["sidetone", 15.13]], [["speech", 35.94], ["music", 13.31], ["outside, rural or natural", 6.25]], [["music", 47.52], ["theremin", 15.54], ["musical instrument", 6.38]], null, null, null, null, [["music", 42.7], ["didgeridoo", 10.29], ["wind instrument, woodwind instrument", 6.0]], null, null, null, null, null, [["livestock, farm animals, working animals", 19.14], ["cattle, bovinae", 15.99], ["moo", 13.49]]], "duration": [0.5, 7.19, 1.99, 2.99, 2.05, 0.45, 1.44, 1.77, 0.58, 0.7, 6.49, 2.1, 0.41, 12.61, 15.54, 5.88, 0.53, 1.86, 33.03, 1.55, 11.2, 0.53, 0.83, 0.96, 1.34, 0.4, 2.06]}

annotations_filtered/g5lJIy5IcWc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 20.31], [23.0, 24.43], [27.0, 27.48], [31.0, 31.19], [33.0, 33.56], [38.0, 37.99], [39.0, 41.15], [47.0, 48.1], [48.0, 52.08], [55.0, 56.2], [57.0, 57.86], [58.0, 58.67], [61.0, 67.61], [72.0, 72.94], [76.0, 81.62], [88.0, 88.38], [88.0, 88.5], [92.0, 92.38], [106.0, 107.15], [114.0, 114.39], [117.0, 120.53], [121.0, 121.51], [123.0, 123.45], [123.0, 125.0]], "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": [54.1, 0.0, 0.0, 0.0, 0.0, 0.0, 57.81, 0.0, 61.97, 0.0, 0.0, 0.0, 70.86, 0.0, 53.4, 0.0, 0.0, 0.0, 0.0, 0.0, 50.46, 0.0, 0.0, 53.91], "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], "duration": [15.31, 1.43, 0.48, 0.19, 0.56, -0.01, 2.15, 1.1, 4.08, 1.2, 0.86, 0.67, 6.61, 0.94, 5.62, 0.38, 0.5, 0.38, 1.15, 0.39, 3.53, 0.51, 0.45, 2.0]}

annotations_filtered/g5m411zcNA4_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 2.74], [4.0, 10.98], [12.0, 15.6], [16.0, 18.45], [19.0, 20.31], [21.0, 22.99], [25.0, 26.22], [27.0, 29.34], [30.0, 31.58], [32.0, 33.84], [35.0, 36.05], [37.0, 37.93], [40.0, 42.65], [43.0, 44.83], [45.0, 46.62], [48.0, 48.78], [51.0, 51.49], [53.0, 53.96], [55.0, 63.86], [67.0, 67.54], [70.0, 70.44], [72.0, 74.41], [75.0, 78.09]], "keep_status": [false, false, false, true, false, false, false, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true], "silence_prob": [0.0, 84.43, 65.44, 46.75, 0.0, 0.0, 0.0, 46.5, 0.0, 0.0, 0.0, 0.0, 38.2, 0.0, 0.0, 0.0, 0.0, 0.0, 58.81, 0.0, 0.0, 66.15, 35.33], "audiomae_on_audioset": [null, null, null, [["hum", 15.06], ["owl", 8.68], ["mains hum", 7.11]], null, null, null, [["music", 37.81], ["gong", 11.17], ["synthesizer", 5.72]], null, null, null, null, [["speech", 53.87], ["radio", 9.98], ["sidetone", 6.54]], null, null, null, null, null, null, null, null, null, [["hum", 24.27], ["sidetone", 16.88], ["whale vocalization", 8.59]]], "duration": [0.74, 6.98, 3.6, 2.45, 1.31, 1.99, 1.22, 2.34, 1.58, 1.84, 1.05, 0.93, 2.65, 1.83, 1.62, 0.78, 0.49, 0.96, 8.86, 0.54, 0.44, 2.41, 3.09]}

annotations_filtered/g6DnsZvudTI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[13.0, 13.76], [17.0, 21.95], [25.0, 28.0], [29.0, 50.06], [53.0, 54.23], [58.0, 61.25]], "keep_status": [false, false, false, false, false, false], "silence_prob": [0.0, 79.94, 33.06, 69.47, 0.0, 72.75], "audiomae_on_audioset": [null, null, [["thunk", 53.01], ["speech", 28.62], ["telephone", 2.35]], null, null, null], "duration": [0.76, 4.95, 3.0, 21.06, 1.23, 3.25]}

annotations_filtered/g6mF_yokyiA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[13.0, 13.83], [25.0, 25.3], [28.0, 29.29], [35.0, 35.19], [41.0, 42.04], [44.0, 44.02], [49.0, 50.45], [51.0, 51.63]], "keep_status": [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], "audiomae_on_audioset": [null, null, null, null, null, null, null, null], "duration": [0.83, 0.3, 1.29, 0.19, 1.04, 0.02, 1.45, 0.63]}

annotations_filtered/g6sSw9vrO0s_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 4.04], [6.0, 22.38], [23.0, 39.51], [43.0, 49.76], [57.0, 57.75], [59.0, 59.1], [70.0, 70.21], [72.0, 72.57]], "keep_status": [false, false, false, false, false, false, false, false], "silence_prob": [0.0, 33.42, 35.91, 53.4, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, [["music", 35.25], ["throbbing", 21.32], ["hum", 19.38]], [["music", 83.44], ["throbbing", 2.29], ["hum", 1.97]], null, null, null, null, null], "duration": [0.04, 16.38, 16.51, 6.76, 0.75, 0.1, 0.21, 0.57]}