WIP

annotations_filtered/e-5SVm2NUe8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 6.34], [14.0, 16.06], [19.0, 19.68], [31.0, 32.22], [37.0, 37.44], [39.0, 39.55], [42.0, 46.14], [48.0, 49.1], [54.0, 54.75], [71.0, 72.67], [75.0, 75.47], [76.0, 78.31], [82.0, 94.93], [96.0, 96.96], [98.0, 98.79], [100.0, 100.65], [103.0, 103.91], [110.0, 110.98], [112.0, 112.56], [117.0, 117.95], [120.0, 119.84], [122.0, 123.06], [124.0, 124.83]], "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], "silence_prob": [0.0, 45.27, 0.0, 0.0, 0.0, 0.0, 85.9, 0.0, 0.0, 0.0, 0.0, 36.18, 54.97, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, [["sidetone", 40.92], ["speech", 15.39], ["radio", 6.02]], null, null, null, null, null, null, null, null, null, [["speech", 57.64], ["sidetone", 8.31], ["radio", 4.27]], null, null, null, null, null, null, null, null, null, null, null], "duration": [1.34, 2.06, 0.68, 1.22, 0.44, 0.55, 4.14, 1.1, 0.75, 1.67, 0.47, 2.31, 12.93, 0.96, 0.79, 0.65, 0.91, 0.98, 0.56, 0.95, -0.16, 1.06, 0.83]}

annotations_filtered/e-6fZ7wiPQk_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[31.0, 38.55], [41.0, 99.42], [100.0, 103.05], [104.0, 108.03], [109.0, 110.69], [112.0, 115.11], [118.0, 129.64], [131.0, 138.43]], "keep_status": [false, false, false, false, false, false, false, true], "silence_prob": [34.46, 0.0, 96.29, 98.8, 0.0, 87.55, 32.29, 32.31], "audiomae_on_audioset": [[["hum", 47.98], ["mains hum", 22.12], ["music", 14.06]], null, null, null, null, null, [["hum", 38.82], ["mains hum", 34.86], ["throbbing", 8.39]], [["noise", 14.56], ["speech", 9.28], ["hum", 6.82]]], "duration": [7.55, 58.42, 3.05, 4.03, 1.69, 3.11, 11.64, 7.43]}

annotations_filtered/e-NMI6o5ynk_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 7.82], [12.0, 13.2], [14.0, 15.67], [16.0, 27.33], [31.0, 32.07], [33.0, 51.56]], "keep_status": [false, false, false, false, false, false], "silence_prob": [35.1, 0.0, 0.0, 33.47, 0.0, 32.65], "audiomae_on_audioset": [[["music", 80.14], ["synthesizer", 1.36], ["sampler", 1.24]], null, null, [["music", 79.07], ["throbbing", 1.48], ["sampler", 1.31]], null, [["music", 71.27], ["singing", 3.02], ["didgeridoo", 2.96]]], "duration": [2.82, 1.2, 1.67, 11.33, 1.07, 18.56]}

annotations_filtered/e-oTHFL5_ec_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 6.59], [12.0, 12.97], [15.0, 15.36], [17.0, 17.34], [19.0, 19.26], [27.0, 27.6], [35.0, 35.95], [37.0, 37.91], [39.0, 39.45], [41.0, 42.47], [46.0, 46.36], [50.0, 50.6], [72.0, 73.77], [74.0, 77.52], [78.0, 78.31], [82.0, 85.21]], "keep_status": [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, 99.78, 0.0, 33.07], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["livestock, farm animals, working animals", 41.31], ["moo", 26.89], ["cattle, bovinae", 16.38]]], "duration": [0.59, 0.97, 0.36, 0.34, 0.26, 0.6, 0.95, 0.91, 0.45, 1.47, 0.36, 0.6, 1.77, 3.52, 0.31, 3.21]}

annotations_filtered/e0HzBST_794_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 1.86], [6.0, 15.9], [18.0, 19.84], [27.0, 34.37], [35.0, 49.94], [66.0, 66.11], [68.0, 69.35], [90.0, 90.17], [94.0, 99.76], [105.0, 112.06], [124.0, 157.06], [157.0, 157.1], [157.0, 157.13], [157.0, 157.17]], "keep_status": [false, false, false, false, false, false, false, false, true, true, false, false, false, false], "silence_prob": [0.0, 31.34, 0.0, 30.99, 32.63, 0.0, 0.0, 0.0, 35.36, 32.15, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, [["livestock, farm animals, working animals", 46.24], ["cattle, bovinae", 21.08], ["moo", 11.12]], null, [["moo", 35.5], ["cattle, bovinae", 33.52], ["livestock, farm animals, working animals", 25.6]], [["animal", 63.47], ["grunt", 17.72], ["livestock, farm animals, working animals", 3.93]], null, null, null, [["music", 22.85], ["whack, thwack", 14.08], ["speech", 13.15]], [["boing", 24.17], ["speech", 15.59], ["music", 14.44]], null, null, null, null], "duration": [-0.14, 9.9, 1.84, 7.37, 14.94, 0.11, 1.35, 0.17, 5.76, 7.06, 33.06, 0.1, 0.13, 0.17]}

annotations_filtered/e0UZ0rc0KH4_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[9.0, 9.88], [10.0, 10.5], [11.0, 11.67], [13.0, 13.1], [13.0, 14.88], [16.0, 23.33], [24.0, 34.57], [37.0, 36.98], [37.0, 37.66], [44.0, 50.04], [52.0, 51.83], [52.0, 56.54], [57.0, 59.61], [61.0, 61.75], [63.0, 63.42], [64.0, 65.31], [66.0, 75.52], [78.0, 80.69], [82.0, 82.73], [86.0, 86.12], [87.0, 87.02], [90.0, 92.18], [93.0, 98.05], [99.0, 99.44], [100.0, 100.84], [104.0, 105.29], [106.0, 115.64], [119.0, 119.65], [120.0, 125.27], [126.0, 130.0], [131.0, 133.02]], "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, 0.0, 0.0, 0.0, 52.56, 57.32, 0.0, 0.0, 58.89, 0.0, 52.45, 60.14, 0.0, 0.0, 0.0, 46.43, 58.89, 0.0, 0.0, 0.0, 77.7, 64.07, 0.0, 0.0, 0.0, 78.89, 0.0, 71.43, 69.2, 63.42], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 48.08], ["hum", 20.56], ["throbbing", 8.19]], null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.88, 0.5, 0.67, 0.1, 1.88, 7.33, 10.57, -0.02, 0.66, 6.04, -0.17, 4.54, 2.61, 0.75, 0.42, 1.31, 9.52, 2.69, 0.73, 0.12, 0.02, 2.18, 5.05, 0.44, 0.84, 1.29, 9.64, 0.65, 5.27, 4.0, 2.02]}

annotations_filtered/e0cR6R3SccI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[27.0, 28.07], [31.0, 50.21]], "keep_status": [false, false], "silence_prob": [0.0, 66.15], "audiomae_on_audioset": [null, null], "duration": [1.07, 19.21]}

annotations_filtered/e0d5svC0xQU_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[7.0, 10.44], [20.0, 20.78], [21.0, 23.23], [26.0, 30.45], [31.0, 31.41], [34.0, 34.05], [44.0, 48.0], [48.0, 48.61], [52.0, 53.92], [55.0, 56.86], [59.0, 58.89], [61.0, 61.74], [64.0, 64.18], [66.0, 67.15]], "keep_status": [true, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [29.77, 0.0, 59.15, 52.39, 0.0, 0.0, 64.86, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [[["music", 38.53], ["throbbing", 10.18], ["hum", 8.84]], null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [3.44, 0.78, 2.23, 4.45, 0.41, 0.05, 4.0, 0.61, 1.92, 1.86, -0.11, 0.74, 0.18, 1.15]}

annotations_filtered/e0fFbNV1ySY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[7.0, 18.67], [19.0, 20.76], [25.0, 25.88], [27.0, 28.38], [29.0, 43.44], [45.0, 45.49], [49.0, 50.53], [52.0, 53.28], [54.0, 54.82], [58.0, 58.4], [60.0, 70.92], [72.0, 75.15], [76.0, 77.95], [80.0, 92.99], [93.0, 96.69]], "keep_status": [false, false, false, false, true, false, false, false, false, false, false, false, false, false, false], "silence_prob": [60.98, 0.0, 0.0, 0.0, 30.33, 0.0, 0.0, 0.0, 0.0, 0.0, 55.74, 71.43, 0.0, 69.07, 80.64], "audiomae_on_audioset": [null, null, null, null, [["sidetone", 48.61], ["gong", 9.73], ["music", 9.58]], null, null, null, null, null, null, null, null, null, null], "duration": [11.67, 1.76, 0.88, 1.38, 14.44, 0.49, 1.53, 1.28, 0.82, 0.4, 10.92, 3.15, 1.95, 12.99, 3.69]}

annotations_filtered/e0qpJCNLViQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 6.86], [8.0, 8.24], [30.0, 30.42], [31.0, 33.52], [36.0, 37.34], [41.0, 42.11], [45.0, 53.86], [61.0, 70.02], [70.0, 71.69], [74.0, 74.7], [76.0, 78.09], [79.0, 80.15], [82.0, 84.5], [85.0, 86.41], [90.0, 90.71], [91.0, 101.02], [101.0, 102.54], [105.0, 106.51], [108.0, 108.48], [109.0, 110.59]], "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, 0.0, 96.54, 0.0, 0.0, 41.12, 41.54, 0.0, 0.0, 100.0, 0.0, 99.96, 0.0, 0.0, 84.07, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, [["music", 38.23], ["speech", 28.7], ["hip hop music", 5.37]], [["sidetone", 47.08], ["music", 30.24], ["speech", 6.44]], null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.86, 0.24, 0.42, 2.52, 1.34, 1.11, 8.86, 9.02, 1.69, 0.7, 2.09, 1.15, 2.5, 1.41, 0.71, 10.02, 1.54, 1.51, 0.48, 1.59]}

annotations_filtered/e198XToyAkk_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 5.1], [35.0, 35.58], [41.0, 41.47], [49.0, 49.72], [52.0, 51.75], [64.0, 65.48], [67.0, 66.93], [68.0, 74.49], [76.0, 75.96], [78.0, 77.95], [95.0, 97.78], [103.0, 103.84], [115.0, 115.52], [136.0, 136.81], [138.0, 138.43], [146.0, 148.22]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 44.07, 0.0, 0.0, 35.91, 0.0, 0.0, 0.0, 0.0, 35.71], "audiomae_on_audioset": [null, null, null, null, null, null, null, [["music", 72.9], ["speech", 11.73], ["wail, moan", 3.73]], null, null, [["music", 77.4], ["musical instrument", 5.51], ["synthesizer", 3.79]], null, null, null, null, [["boing", 36.42], ["speech", 18.1], ["ding", 6.88]]], "duration": [0.1, 0.58, 0.47, 0.72, -0.25, 1.48, -0.07, 6.49, -0.04, -0.05, 2.78, 0.84, 0.52, 0.81, 0.43, 2.22]}

annotations_filtered/e1DnltskkWk_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[11.0, 11.67], [45.0, 46.68], [57.0, 58.06], [59.0, 59.64], [62.0, 62.36], [63.0, 63.34], [66.0, 66.55], [68.0, 68.93], [74.0, 75.12], [78.0, 79.96], [86.0, 86.66], [87.0, 87.24], [88.0, 89.06], [96.0, 97.31], [99.0, 99.84], [104.0, 105.33], [106.0, 106.66], [108.0, 109.56], [113.0, 114.2], [117.0, 117.27], [127.0, 127.13], [128.0, 130.45], [131.0, 130.94]], "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, 79.76, 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": [0.67, 1.68, 1.06, 0.64, 0.36, 0.34, 0.55, 0.93, 1.12, 1.96, 0.66, 0.24, 1.06, 1.31, 0.84, 1.33, 0.66, 1.56, 1.2, 0.27, 0.13, 2.45, -0.06]}

annotations_filtered/e1FWoMLjAU0_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[60.0, 61.79], [63.0, 63.86], [65.0, 65.97], [71.0, 72.82], [73.0, 74.76], [76.0, 77.41], [79.0, 81.35], [83.0, 84.21], [86.0, 86.54], [88.0, 90.14], [90.0, 93.43], [94.0, 95.3], [97.0, 97.43]], "keep_status": [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, 94.52, 0.0, 0.0, 77.03, 50.61, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [1.79, 0.86, 0.97, 1.82, 1.76, 1.41, 2.35, 1.21, 0.54, 2.14, 3.43, 1.3, 0.43]}

annotations_filtered/e1bMBM_rgwQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[12.0, 13.32], [14.0, 15.79], [18.0, 19.4], [20.0, 21.63], [26.0, 26.33], [31.0, 31.68], [36.0, 36.34], [42.0, 46.47], [52.0, 53.08], [54.0, 54.65], [57.0, 57.69], [58.0, 59.0], [60.0, 62.26], [67.0, 69.62], [71.0, 71.85]], "keep_status": [false, false, false, false, false, false, false, true, false, false, false, false, true, true, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 47.16, 0.0, 0.0, 0.0, 0.0, 45.92, 46.83, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, [["music", 18.86], ["radio", 10.34], ["noise", 9.59]], null, null, null, null, [["hum", 26.18], ["throbbing", 14.56], ["mains hum", 13.05]], [["speech", 21.24], ["hum", 20.28], ["radio", 17.07]], null], "duration": [1.32, 1.79, 1.4, 1.63, 0.33, 0.68, 0.34, 4.47, 1.08, 0.65, 0.69, 1.0, 2.26, 2.62, 0.85]}

annotations_filtered/e2-VnrdN_Ng_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 2.05], [3.0, 3.0], [4.0, 4.08], [8.0, 7.69], [16.0, 17.02], [18.0, 19.4], [23.0, 23.36], [24.0, 25.46], [34.0, 36.78], [41.0, 42.62], [47.0, 59.29], [60.0, 122.72]], "keep_status": [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, 94.66, 0.0, 67.38, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.05, 0.0, 0.08, -0.31, 1.02, 1.4, 0.36, 1.46, 2.78, 1.62, 12.29, 62.72]}

annotations_filtered/e2FdM0YQSHk_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[13.0, 13.37], [25.0, 26.1], [28.0, 28.26], [30.0, 39.5], [41.0, 63.61]], "keep_status": [false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 92.31, 54.23], "audiomae_on_audioset": [null, null, null, null, null], "duration": [0.37, 1.1, 0.26, 9.5, 22.61]}

annotations_filtered/e2Odq49gEbs_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[16.0, 18.28], [24.0, 38.79], [47.0, 47.02], [53.0, 54.04], [55.0, 55.09], [56.0, 61.87], [64.0, 72.6], [74.0, 75.41], [82.0, 92.38], [94.0, 98.05], [100.0, 107.42], [108.0, 108.89], [110.0, 110.83], [112.0, 112.33], [114.0, 114.67], [115.0, 117.61], [119.0, 120.43], [131.0, 131.43]], "keep_status": [false, true, false, false, false, true, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [48.35, 46.15, 0.0, 0.0, 0.0, 40.16, 40.9, 0.0, 51.28, 73.06, 73.67, 0.0, 0.0, 0.0, 0.0, 55.18, 0.0, 0.0], "audiomae_on_audioset": [[["music", 47.47], ["hum", 21.39], ["mains hum", 7.8]], [["noise", 31.02], ["hum", 27.41], ["music", 10.65]], null, null, null, [["music", 37.27], ["speech", 24.44], ["synthesizer", 6.19]], [["music", 53.75], ["throbbing", 11.1], ["hum", 8.26]], null, null, null, null, null, null, null, null, null, null, null], "duration": [2.28, 14.79, 0.02, 1.04, 0.09, 5.87, 8.6, 1.41, 10.38, 4.05, 7.42, 0.89, 0.83, 0.33, 0.67, 2.61, 1.43, 0.43]}

annotations_filtered/e2RQwVyRSGU_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[26.0, 27.78], [31.0, 32.09], [38.0, 39.11], [40.0, 40.27], [42.0, 43.02], [45.0, 48.59], [65.0, 65.57], [67.0, 99.01]], "keep_status": [false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null], "duration": [1.78, 1.09, 1.11, 0.27, 1.02, 3.59, 0.57, 32.01]}

annotations_filtered/e2dAiCB6igE_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[13.0, 14.0], [16.0, 16.44], [22.0, 22.64], [24.0, 24.75], [28.0, 29.44], [30.0, 31.21], [33.0, 34.42], [39.0, 40.61], [52.0, 52.44], [57.0, 57.84], [61.0, 64.77], [67.0, 69.38], [71.0, 70.87], [72.0, 73.79], [82.0, 88.25], [91.0, 91.84], [93.0, 93.68], [101.0, 107.86]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, false, false, true], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 70.72, 99.91, 0.0, 0.0, 42.67, 0.0, 0.0, 30.25], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["music", 43.51], ["fly, housefly", 8.75], ["buzz", 3.29]], null, null, [["music", 30.06], ["fly, housefly", 7.36], ["mosquito", 6.73]]], "duration": [1.0, 0.44, 0.64, 0.75, 1.44, 1.21, 1.42, 1.61, 0.44, 0.84, 3.77, 2.38, -0.13, 1.79, 6.25, 0.84, 0.68, 6.86]}

annotations_filtered/e2g4Wr81rz8_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 3.3], [10.0, 10.66], [24.0, 24.53], [28.0, 28.21], [33.0, 33.49], [51.0, 54.28], [91.0, 93.23], [94.0, 103.82], [111.0, 116.94], [129.0, 131.94]], "keep_status": [false, false, false, false, false, false, false, false, true, true], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 99.52, 98.27, 99.93, 34.37, 37.09], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, [["speech", 25.95], ["cattle, bovinae", 14.05], ["moo", 12.65]], [["speech", 26.35], ["cattle, bovinae", 11.16], ["moo", 8.22]]], "duration": [1.3, 0.66, 0.53, 0.21, 0.49, 3.28, 2.23, 9.82, 5.94, 2.94]}

annotations_filtered/e2xiwiWd_sM_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 1.22], [2.0, 13.12], [14.0, 14.23], [15.0, 15.89], [17.0, 18.42], [19.0, 20.28], [22.0, 22.27], [23.0, 25.19], [31.0, 31.02], [32.0, 38.21], [40.0, 41.44], [42.0, 44.68], [47.0, 48.3], [49.0, 51.81], [53.0, 54.11], [54.0, 55.22], [56.0, 57.72]], "keep_status": [false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, false], "silence_prob": [0.0, 80.46, 0.0, 0.0, 0.0, 0.0, 0.0, 95.64, 0.0, 48.39, 0.0, 97.0, 0.0, 76.04, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, [["hum", 20.91], ["door", 9.88], ["noise", 8.75]], null, null, null, null, null, null, null], "duration": [0.22, 11.12, 0.23, 0.89, 1.42, 1.28, 0.27, 2.19, 0.02, 6.21, 1.44, 2.68, 1.3, 2.81, 1.11, 1.22, 1.72]}

annotations_filtered/e3BZn-XJ4mU_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[23.0, 35.53], [36.0, 35.58], [36.0, 44.9], [46.0, 49.57], [50.0, 73.52], [75.0, 74.87], [76.0, 78.11], [80.0, 93.9], [97.0, 97.77], [100.0, 111.28], [112.0, 114.76], [117.0, 120.88], [121.0, 129.71], [130.0, 134.22], [137.0, 142.28], [143.0, 143.77]], "keep_status": [true, false, false, true, false, false, true, true, false, false, true, true, true, true, true, false], "silence_prob": [28.15, 0.0, 31.04, 31.14, 29.69, 0.0, 29.67, 27.86, 0.0, 28.48, 29.48, 29.45, 29.51, 28.81, 29.46, 0.0], "audiomae_on_audioset": [[["explosion", 20.85], ["music", 10.57], ["whack, thwack", 7.35]], null, [["speech", 66.87], ["music", 7.8], ["breaking", 2.28]], [["music", 52.39], ["hum", 2.64], ["vehicle", 2.21]], [["livestock, farm animals, working animals", 40.85], ["cattle, bovinae", 28.52], ["moo", 9.5]], null, [["sound effect", 13.32], ["whale vocalization", 12.86], ["roar", 8.7]], [["explosion", 15.17], ["rumble", 14.42], ["music", 6.74]], null, [["music", 72.89], ["throbbing", 6.81], ["hum", 6.34]], [["speech", 10.31], ["hum", 9.14], ["whack, thwack", 8.64]], [["music", 28.39], ["whack, thwack", 8.72], ["throbbing", 8.18]], [["music", 36.92], ["whack, thwack", 9.25], ["clang", 8.34]], [["music", 54.73], ["speech", 7.26], ["boing", 5.89]], [["ding", 31.02], ["boing", 11.6], ["music", 11.59]], null], "duration": [12.53, -0.42, 8.9, 3.57, 23.52, -0.13, 2.11, 13.9, 0.77, 11.28, 2.76, 3.88, 8.71, 4.22, 5.28, 0.77]}

annotations_filtered/e3mwmwPhr08_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[28.0, 30.7], [32.0, 33.52], [36.0, 36.46], [37.0, 38.03], [41.0, 41.77], [44.0, 45.06], [51.0, 51.31], [52.0, 54.55], [57.0, 58.67], [60.0, 61.26], [65.0, 67.37], [68.0, 71.58], [73.0, 75.64], [79.0, 79.47], [80.0, 82.09], [84.0, 86.49], [89.0, 90.59], [95.0, 95.32], [98.0, 98.9], [102.0, 102.91], [104.0, 105.92], [106.0, 108.3], [110.0, 112.65], [114.0, 115.43], [119.0, 121.9], [127.0, 130.38], [131.0, 131.9], [133.0, 133.89], [136.0, 136.68], [138.0, 139.04], [141.0, 143.94], [145.0, 146.72], [148.0, 151.33], [153.0, 156.68]], "keep_status": [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, true, false, false, false], "silence_prob": [75.07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 0.0, 100.0, 98.1, 94.95, 0.0, 92.8, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 44.01, 100.0, 0.0, 64.75, 57.09, 0.0, 0.0, 0.0, 0.0, 49.68, 0.0, 56.55, 60.89], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["speech", 37.67], ["sine wave", 22.69], ["singing bowl", 7.04]], null, null, null, null, null, null, null, null, [["gong", 30.58], ["music", 28.16], ["singing bowl", 10.96]], null, null, null], "duration": [2.7, 1.52, 0.46, 1.03, 0.77, 1.06, 0.31, 2.55, 1.67, 1.26, 2.37, 3.58, 2.64, 0.47, 2.09, 2.49, 1.59, 0.32, 0.9, 0.91, 1.92, 2.3, 2.65, 1.43, 2.9, 3.38, 0.9, 0.89, 0.68, 1.04, 2.94, 1.72, 3.33, 3.68]}

annotations_filtered/e4-STTC0pNc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[34.0, 39.83], [41.0, 64.45], [67.0, 75.76], [78.0, 78.44], [80.0, 81.06], [82.0, 87.96], [89.0, 101.93], [102.0, 102.17], [106.0, 109.83]], "keep_status": [false, false, false, false, false, false, false, false, false], "silence_prob": [69.07, 75.72, 99.99, 0.0, 0.0, 99.97, 69.61, 0.0, 41.91], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, [["music", 63.15], ["theremin", 16.52], ["ambient music", 6.96]]], "duration": [5.83, 23.45, 8.76, 0.44, 1.06, 5.96, 12.93, 0.17, 3.83]}

annotations_filtered/e434wHYilA0_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 9.04], [9.0, 10.72], [13.0, 13.05], [18.0, 25.19], [27.0, 27.51], [28.0, 28.29], [29.0, 30.4], [32.0, 33.93], [35.0, 35.16], [36.0, 47.36], [48.0, 48.68], [50.0, 57.35], [60.0, 67.63], [68.0, 69.16], [71.0, 71.39], [75.0, 75.0], [77.0, 77.85], [79.0, 82.68], [87.0, 88.06], [89.0, 89.78], [90.0, 89.85], [90.0, 93.93], [94.0, 104.55], [106.0, 105.93], [108.0, 108.75], [109.0, 110.22], [111.0, 111.81], [112.0, 118.12], [121.0, 130.32], [132.0, 147.88]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, true, false, false, false, true, false, false, false, false, false, true, true, false], "silence_prob": [58.22, 0.0, 0.0, 46.86, 0.0, 0.0, 0.0, 0.0, 0.0, 45.88, 0.0, 40.24, 45.08, 0.0, 0.0, 0.0, 0.0, 43.05, 0.0, 0.0, 0.0, 44.32, 40.88, 0.0, 0.0, 0.0, 0.0, 33.28, 32.46, 29.75], "audiomae_on_audioset": [null, null, null, [["speech", 60.94], ["radio", 31.31], ["hum", 1.04]], null, null, null, null, null, [["mains hum", 59.72], ["hum", 24.53], ["speech", 8.86]], null, [["speech", 73.11], ["radio", 8.72], ["mains hum", 3.79]], [["music", 25.7], ["hum", 19.16], ["mains hum", 14.47]], null, null, null, null, [["music", 40.55], ["hum", 11.37], ["didgeridoo", 7.63]], null, null, null, [["music", 17.28], ["radio", 14.73], ["hum", 12.96]], [["speech", 35.86], ["hum", 23.66], ["mains hum", 22.62]], null, null, null, null, [["music", 19.86], ["speech", 12.88], ["cough", 9.91]], [["music", 36.19], ["speech", 26.44], ["didgeridoo", 4.66]], [["hum", 48.66], ["mains hum", 31.37], ["throbbing", 6.21]]], "duration": [7.04, 1.72, 0.05, 7.19, 0.51, 0.29, 1.4, 1.93, 0.16, 11.36, 0.68, 7.35, 7.63, 1.16, 0.39, 0.0, 0.85, 3.68, 1.06, 0.78, -0.15, 3.93, 10.55, -0.07, 0.75, 1.22, 0.81, 6.12, 9.32, 15.88]}

annotations_filtered/e48T01eVXtU_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 11.52], [12.0, 12.23], [21.0, 22.49], [37.0, 44.41], [48.0, 49.92], [52.0, 56.61], [57.0, 58.36], [65.0, 68.84], [73.0, 74.48], [75.0, 75.22], [76.0, 81.7], [91.0, 91.47], [92.0, 94.53], [96.0, 99.2], [102.0, 103.32], [104.0, 108.65], [109.0, 108.9], [110.0, 110.17], [113.0, 114.52], [117.0, 118.47]], "keep_status": [false, false, false, false, false, false, false, false, false, false, true, false, true, true, false, true, false, false, false, false], "silence_prob": [38.15, 0.0, 0.0, 30.13, 0.0, 30.45, 0.0, 29.71, 0.0, 0.0, 30.93, 0.0, 30.46, 32.8, 0.0, 30.68, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [[["livestock, farm animals, working animals", 43.41], ["cattle, bovinae", 28.68], ["moo", 14.11]], null, null, [["music", 42.17], ["speech", 24.34], ["cattle, bovinae", 5.57]], null, [["moo", 42.12], ["cattle, bovinae", 26.57], ["livestock, farm animals, working animals", 9.8]], null, [["music", 59.82], ["speech", 17.54], ["boing", 5.08]], null, null, [["fly, housefly", 24.32], ["mosquito", 18.31], ["music", 16.1]], null, [["music", 45.94], ["boing", 12.05], ["speech", 5.94]], [["music", 40.04], ["speech", 9.65], ["musical instrument", 9.28]], null, [["cattle, bovinae", 23.48], ["moo", 21.46], ["fart", 11.48]], null, null, null, null], "duration": [7.52, 0.23, 1.49, 7.41, 1.92, 4.61, 1.36, 3.84, 1.48, 0.22, 5.7, 0.47, 2.53, 3.2, 1.32, 4.65, -0.1, 0.17, 1.52, 1.47]}

annotations_filtered/e4Dlc6yqJuA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 12.73], [14.0, 16.58], [18.0, 42.06], [45.0, 45.47], [47.0, 48.05]], "keep_status": [false, false, true, false, false], "silence_prob": [90.43, 89.9, 31.58, 0.0, 0.0], "audiomae_on_audioset": [null, null, [["music", 40.2], ["hum", 19.74], ["mains hum", 7.03]], null, null], "duration": [4.73, 2.58, 24.06, 0.47, 1.05]}

annotations_filtered/e4RGh5iAykY_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 11.28], [13.0, 13.61], [16.0, 16.93], [18.0, 17.81], [20.0, 20.76], [23.0, 24.06], [26.0, 29.47], [30.0, 30.7], [34.0, 34.72], [38.0, 41.05], [43.0, 42.84], [45.0, 45.01], [48.0, 48.24], [51.0, 51.7], [55.0, 63.58], [64.0, 64.88], [66.0, 74.43], [75.0, 76.03], [86.0, 86.36], [87.0, 88.48], [89.0, 89.65], [90.0, 94.39], [108.0, 108.33], [115.0, 115.3], [117.0, 117.51], [118.0, 119.01], [121.0, 121.37], [123.0, 125.2], [127.0, 127.6], [128.0, 134.67], [136.0, 136.49], [137.0, 139.14], [143.0, 143.11], [144.0, 149.15], [150.0, 151.24]], "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], "silence_prob": [38.06, 0.0, 0.0, 0.0, 0.0, 0.0, 99.87, 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 98.93, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 100.0, 0.0, 100.0, 0.0, 100.0, 0.0], "audiomae_on_audioset": [[["music", 56.5], ["theremin", 14.16], ["musical instrument", 4.94]], 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": [10.28, 0.61, 0.93, -0.19, 0.76, 1.06, 3.47, 0.7, 0.72, 3.05, -0.16, 0.01, 0.24, 0.7, 8.58, 0.88, 8.43, 1.03, 0.36, 1.48, 0.65, 4.39, 0.33, 0.3, 0.51, 1.01, 0.37, 2.2, 0.6, 6.67, 0.49, 2.14, 0.11, 5.15, 1.24]}

annotations_filtered/e4yCxKv-2qw_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[0.0, 0.24], [6.0, 8.77], [11.0, 12.46], [14.0, 16.16], [20.0, 22.86], [24.0, 28.93], [30.0, 31.56], [33.0, 35.16], [36.0, 36.58], [37.0, 39.33], [41.0, 41.64], [45.0, 44.98], [46.0, 48.22], [50.0, 51.75], [52.0, 53.65], [55.0, 55.59], [59.0, 59.61], [60.0, 60.45], [61.0, 61.05], [67.0, 67.61], [68.0, 67.64], [70.0, 71.0], [76.0, 75.98], [77.0, 77.06], [78.0, 80.84], [82.0, 84.15], [85.0, 86.56], [88.0, 88.4], [89.0, 90.31], [91.0, 93.63], [97.0, 99.54], [101.0, 101.09], [102.0, 103.1], [104.0, 104.5], [106.0, 106.79], [110.0, 110.37], [111.0, 112.31], [114.0, 114.3], [116.0, 118.39]], "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, true, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 32.09, 0.0, 36.08, 72.75, 68.41, 0.0, 52.86, 0.0, 50.61, 0.0, 0.0, 69.34, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 79.59, 41.89, 0.0, 0.0, 0.0, 72.16, 60.51, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 77.7], "audiomae_on_audioset": [null, [["sound effect", 15.56], ["dishes, pots, and pans", 10.82], ["grunt", 7.11]], null, [["creak", 21.53], ["speech", 20.32], ["tick-tock", 17.06]], null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, [["whale vocalization", 39.02], ["sheep", 9.53], ["bleat", 7.17]], null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.24, 2.77, 1.46, 2.16, 2.86, 4.93, 1.56, 2.16, 0.58, 2.33, 0.64, -0.02, 2.22, 1.75, 1.65, 0.59, 0.61, 0.45, 0.05, 0.61, -0.36, 1.0, -0.02, 0.06, 2.84, 2.15, 1.56, 0.4, 1.31, 2.63, 2.54, 0.09, 1.1, 0.5, 0.79, 0.37, 1.31, 0.3, 2.39]}

annotations_filtered/e56FOw5rIhw_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[50.0, 50.28], [51.0, 77.95]], "keep_status": [false, false], "silence_prob": [0.0, 30.83], "audiomae_on_audioset": [null, [["music", 56.45], ["throbbing", 10.65], ["musical instrument", 4.5]]], "duration": [0.28, 26.95]}

annotations_filtered/e5LZR3vCkzo_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 8.13], [10.0, 10.88], [13.0, 14.5], [16.0, 20.17], [27.0, 64.5], [66.0, 66.46], [70.0, 78.75], [79.0, 80.33], [81.0, 108.63]], "keep_status": [false, false, false, true, false, false, false, false, true], "silence_prob": [60.42, 0.0, 0.0, 35.0, 0.0, 0.0, 32.52, 0.0, 30.57], "audiomae_on_audioset": [null, null, null, [["music", 36.42], ["throbbing", 16.46], ["hum", 16.37]], null, null, [["music", 68.46], ["speech", 5.44], ["throbbing", 4.54]], null, [["music", 42.24], ["speech", 12.1], ["whack, thwack", 10.14]]], "duration": [3.13, 0.88, 1.5, 4.17, 37.5, 0.46, 8.75, 1.33, 27.63]}

annotations_filtered/e5SbxMFk6Vo_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 1.88], [7.0, 7.6], [9.0, 10.89], [23.0, 23.4], [27.0, 27.14], [37.0, 36.8], [50.0, 50.48], [53.0, 53.33], [57.0, 60.72], [64.0, 67.34], [70.0, 69.72], [70.0, 69.75], [70.0, 79.74], [84.0, 94.56], [103.0, 102.78], [103.0, 118.83], [120.0, 120.06], [126.0, 126.3]], "keep_status": [false, false, false, false, false, false, false, false, false, true, false, false, false, true, false, true, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 31.3, 30.29, 0.0, 0.0, 30.23, 30.0, 0.0, 30.64, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, [["music", 55.26], ["speech", 15.52], ["electronic music", 3.52]], [["music", 54.58], ["speech", 9.73], ["boing", 4.41]], null, null, [["music", 74.44], ["speech", 10.08], ["didgeridoo", 4.33]], [["music", 55.45], ["brass instrument", 8.85], ["trombone", 4.94]], null, [["livestock, farm animals, working animals", 31.77], ["music", 18.64], ["moo", 17.12]], null, null], "duration": [0.88, 0.6, 1.89, 0.4, 0.14, -0.2, 0.48, 0.33, 3.72, 3.34, -0.28, -0.25, 9.74, 10.56, -0.22, 15.83, 0.06, 0.3]}

annotations_filtered/e5Tb2YSkGh0_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 5.29], [7.0, 7.5], [10.0, 10.49], [15.0, 15.36], [26.0, 27.8], [28.0, 28.95], [29.0, 29.67], [40.0, 67.64]], "keep_status": [false, false, false, false, false, false, false, true], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 34.55], "audiomae_on_audioset": [null, null, null, null, null, null, null, [["radio", 46.5], ["noise", 8.36], ["sidetone", 5.75]]], "duration": [0.29, 0.5, 0.49, 0.36, 1.8, 0.95, 0.67, 27.64]}

annotations_filtered/e5cg1EeFISo_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 8.04], [9.0, 9.07], [30.0, 30.27], [39.0, 40.47], [50.0, 50.77], [74.0, 74.6], [76.0, 75.83], [91.0, 91.03], [103.0, 106.86], [109.0, 109.76], [127.0, 127.01], [128.0, 128.33]], "keep_status": [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, 51.5, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.04, 0.07, 0.27, 1.47, 0.77, 0.6, -0.17, 0.03, 3.86, 0.76, 0.01, 0.33]}

annotations_filtered/e6B-T4KYIFQ_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[6.0, 7.64], [8.0, 29.32], [32.0, 33.89], [36.0, 38.58], [39.0, 38.75], [39.0, 39.04], [39.0, 49.54], [54.0, 56.76], [58.0, 59.64], [63.0, 73.3], [77.0, 98.63], [103.0, 105.61], [106.0, 118.56], [126.0, 126.32], [127.0, 143.04]], "keep_status": [false, true, false, true, false, false, false, false, false, false, false, true, false, false, false], "silence_prob": [0.0, 49.04, 0.0, 46.19, 0.0, 0.0, 48.74, 94.22, 0.0, 47.78, 35.4, 30.67, 33.5, 0.0, 39.91], "audiomae_on_audioset": [null, [["music", 32.85], ["speech", 9.39], ["boing", 8.27]], null, [["speech", 20.76], ["music", 6.79], ["singing bowl", 3.8]], null, null, [["music", 35.7], ["speech", 29.53], ["whack, thwack", 6.13]], null, null, [["music", 66.71], ["electronic music", 4.64], ["thunk", 3.67]], [["music", 60.17], ["electronic music", 6.74], ["ambient music", 5.88]], [["speech", 32.66], ["sheep", 22.94], ["bleat", 13.85]], [["music", 71.56], ["scary music", 6.47], ["speech", 4.28]], null, [["speech", 58.1], ["animal", 14.81], ["dog", 5.38]]], "duration": [1.64, 21.32, 1.89, 2.58, -0.25, 0.04, 10.54, 2.76, 1.64, 10.3, 21.63, 2.61, 12.56, 0.32, 16.04]}

annotations_filtered/e6WyN4z0VGc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[1.0, 3.23], [5.0, 6.02], [6.0, 8.8], [9.0, 11.65], [13.0, 14.35], [15.0, 17.73], [21.0, 26.77], [27.0, 29.0], [30.0, 34.89], [36.0, 36.95], [40.0, 41.76], [45.0, 48.1], [50.0, 54.45], [56.0, 58.55], [59.0, 63.37], [64.0, 65.18], [66.0, 66.55], [69.0, 70.56], [72.0, 73.06], [74.0, 85.06], [86.0, 96.03], [98.0, 98.49], [101.0, 102.84], [105.0, 105.46], [108.0, 108.55], [109.0, 110.79], [113.0, 113.98], [117.0, 117.96], [122.0, 123.15], [124.0, 124.63], [130.0, 130.1], [131.0, 132.26], [135.0, 138.28], [141.0, 141.61], [145.0, 144.71], [147.0, 147.6]], "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], "silence_prob": [90.43, 0.0, 91.3, 99.68, 0.0, 100.0, 100.0, 86.09, 100.0, 0.0, 0.0, 99.97, 100.0, 81.35, 100.0, 0.0, 0.0, 0.0, 0.0, 99.1, 59.51, 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], "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], "duration": [2.23, 1.02, 2.8, 2.65, 1.35, 2.73, 5.77, 2.0, 4.89, 0.95, 1.76, 3.1, 4.45, 2.55, 4.37, 1.18, 0.55, 1.56, 1.06, 11.06, 10.03, 0.49, 1.84, 0.46, 0.55, 1.79, 0.98, 0.96, 1.15, 0.63, 0.1, 1.26, 3.28, 0.61, -0.29, 0.6]}

annotations_filtered/e6omnFC9jC4_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 2.69], [6.0, 7.05], [14.0, 14.86], [16.0, 21.66], [23.0, 26.03], [27.0, 28.07], [29.0, 44.91], [46.0, 47.01], [48.0, 48.49], [54.0, 57.3], [58.0, 59.36], [60.0, 66.45], [68.0, 71.34], [72.0, 73.16], [74.0, 74.04], [74.0, 78.56], [81.0, 82.58], [84.0, 86.27], [87.0, 88.23], [91.0, 92.67], [94.0, 96.47], [99.0, 100.5], [103.0, 104.18], [105.0, 106.24], [109.0, 111.82], [112.0, 113.04], [114.0, 114.59], [116.0, 118.74], [122.0, 123.33]], "keep_status": [false, false, false, true, true, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 45.4, 47.78, 0.0, 62.78, 0.0, 0.0, 79.94, 0.0, 80.64, 64.29, 0.0, 0.0, 37.25, 0.0, 61.37, 0.0, 0.0, 70.86, 0.0, 0.0, 0.0, 84.62, 0.0, 0.0, 54.7, 0.0], "audiomae_on_audioset": [null, null, null, [["eruption", 23.02], ["hum", 15.35], ["explosion", 12.45]], [["hum", 17.43], ["mains hum", 10.7], ["static", 9.42]], null, null, null, null, null, null, null, null, null, null, [["fly, housefly", 32.56], ["insect", 25.9], ["fart", 7.16]], null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.69, 1.05, 0.86, 5.66, 3.03, 1.07, 15.91, 1.01, 0.49, 3.3, 1.36, 6.45, 3.34, 1.16, 0.04, 4.56, 1.58, 2.27, 1.23, 1.67, 2.47, 1.5, 1.18, 1.24, 2.82, 1.04, 0.59, 2.74, 1.33]}

annotations_filtered/e72atw7hctA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[9.0, 9.12], [19.0, 20.78], [25.0, 25.49], [26.0, 26.4], [31.0, 40.66], [43.0, 43.75], [46.0, 53.52], [58.0, 58.95], [72.0, 79.02], [83.0, 85.4]], "keep_status": [false, false, false, false, false, false, true, false, false, false], "silence_prob": [0.0, 0.0, 0.0, 0.0, 54.7, 0.0, 38.78, 0.0, 48.56, 53.04], "audiomae_on_audioset": [null, null, null, null, null, null, [["music", 40.44], ["didgeridoo", 19.35], ["throbbing", 5.95]], null, [["throbbing", 53.35], ["hum", 32.86], ["music", 5.16]], null], "duration": [0.12, 1.78, 0.49, 0.4, 9.66, 0.75, 7.52, 0.95, 7.02, 2.4]}

annotations_filtered/e77ybggTHHg_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 9.21], [12.0, 13.86], [18.0, 18.28], [23.0, 28.98], [30.0, 37.56], [40.0, 43.17], [46.0, 59.22], [64.0, 68.35], [71.0, 72.54], [76.0, 79.24], [81.0, 82.44], [83.0, 85.8], [88.0, 90.75], [93.0, 107.01], [108.0, 110.29], [112.0, 122.89], [125.0, 126.59]], "keep_status": [false, false, false, false, false, false, false, true, false, false, false, false, true, false, true, false, false], "silence_prob": [30.69, 0.0, 0.0, 35.87, 37.54, 33.65, 31.21, 35.39, 0.0, 34.07, 0.0, 37.83, 34.13, 32.55, 36.37, 32.49, 0.0], "audiomae_on_audioset": [[["hum", 41.09], ["throbbing", 23.24], ["music", 14.98]], null, null, [["music", 40.74], ["hum", 26.2], ["throbbing", 18.64]], [["music", 29.49], ["hum", 27.32], ["throbbing", 19.42]], [["music", 57.65], ["throbbing", 13.03], ["hum", 8.71]], [["music", 69.6], ["throbbing", 11.09], ["hum", 9.4]], [["music", 42.46], ["throbbing", 11.6], ["hum", 11.21]], null, [["music", 69.12], ["throbbing", 6.56], ["hum", 2.56]], null, [["music", 59.93], ["electronic music", 9.56], ["throbbing", 8.96]], [["music", 28.23], ["speech", 17.28], ["hum", 8.43]], [["music", 64.37], ["throbbing", 12.02], ["hum", 5.09]], [["music", 58.11], ["didgeridoo", 5.22], ["synthesizer", 3.89]], [["throbbing", 31.81], ["music", 26.61], ["hum", 23.37]], null], "duration": [7.21, 1.86, 0.28, 5.98, 7.56, 3.17, 13.22, 4.35, 1.54, 3.24, 1.44, 2.8, 2.75, 14.01, 2.29, 10.89, 1.59]}

annotations_filtered/e7Efjj3_uME_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[8.0, 17.46], [38.0, 39.18], [50.0, 50.19], [60.0, 60.05], [64.0, 65.01], [72.0, 73.18], [74.0, 74.29], [79.0, 79.15], [80.0, 80.23], [83.0, 84.4], [85.0, 85.67], [88.0, 101.92], [103.0, 103.03], [104.0, 104.62], [106.0, 106.05], [108.0, 108.95], [113.0, 113.48]], "keep_status": [true, false, false, false, false, false, false, false, false, false, false, true, false, false, false, false, false], "silence_prob": [31.83, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 36.12, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [[["didgeridoo", 10.6], ["hum", 10.19], ["music", 8.2]], null, null, null, null, null, null, null, null, null, null, [["livestock, farm animals, working animals", 27.78], ["moo", 21.77], ["cattle, bovinae", 20.36]], null, null, null, null, null], "duration": [9.46, 1.18, 0.19, 0.05, 1.01, 1.18, 0.29, 0.15, 0.23, 1.4, 0.67, 13.92, 0.03, 0.62, 0.05, 0.95, 0.48]}

annotations_filtered/e7X01_j_oDA_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 6.42], [7.0, 17.42], [19.0, 28.41], [29.0, 30.48], [33.0, 33.07], [34.0, 39.8], [40.0, 40.86], [43.0, 46.16], [48.0, 49.49], [52.0, 54.89], [55.0, 65.08], [67.0, 74.49], [76.0, 80.32], [81.0, 86.53], [95.0, 95.88], [98.0, 101.49], [102.0, 104.28], [106.0, 111.0], [113.0, 115.75]], "keep_status": [false, true, false, false, false, false, false, false, false, true, false, false, true, false, false, false, false, false, false], "silence_prob": [63.96, 32.11, 81.89, 0.0, 0.0, 57.81, 0.0, 34.57, 0.0, 38.42, 71.72, 38.26, 31.92, 99.4, 0.0, 53.65, 48.91, 52.27, 53.4], "audiomae_on_audioset": [null, [["music", 41.95], ["hum", 7.93], ["synthesizer", 7.74]], null, null, null, null, null, [["sidetone", 67.43], ["speech", 13.52], ["boing", 11.32]], null, [["noise", 27.02], ["speech", 24.35], ["radio", 17.16]], null, [["music", 62.41], ["musical instrument", 9.22], ["synthesizer", 7.75]], [["music", 34.26], ["sound effect", 6.29], ["cacophony", 2.92]], null, null, null, [["music", 60.43], ["synthesizer", 8.61], ["musical instrument", 5.81]], null, null], "duration": [2.42, 10.42, 9.41, 1.48, 0.07, 5.8, 0.86, 3.16, 1.49, 2.89, 10.08, 7.49, 4.32, 5.53, 0.88, 3.49, 2.28, 5.0, 2.75]}

annotations_filtered/e7dbge0Zk5A_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[43.0, 79.83], [81.0, 95.72], [96.0, 96.79], [97.0, 100.25], [100.0, 111.86], [113.0, 115.32], [116.0, 116.48], [118.0, 118.44], [124.0, 124.92]], "keep_status": [false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 35.1, 0.0, 69.61, 50.41, 92.97, 0.0, 0.0, 0.0], "audiomae_on_audioset": [null, [["speech", 64.32], ["buzz", 7.43], ["electric shaver, electric razor", 3.78]], null, null, null, null, null, null, null], "duration": [36.83, 14.72, 0.79, 3.25, 11.86, 2.32, 0.48, 0.44, 0.92]}

annotations_filtered/e7qjmOp9G34_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 51.98], [54.0, 80.22], [81.0, 85.62], [87.0, 88.48], [90.0, 90.22], [96.0, 97.8], [99.0, 107.5], [110.0, 111.82], [115.0, 125.29], [127.0, 130.92], [131.0, 136.7], [137.0, 139.85], [140.0, 141.66]], "keep_status": [false, false, false, false, false, false, false, false, false, true, true, true, false], "silence_prob": [0.0, 32.25, 37.76, 0.0, 0.0, 0.0, 30.35, 0.0, 29.97, 40.45, 40.59, 44.01, 0.0], "audiomae_on_audioset": [null, [["music", 38.4], ["throbbing", 25.24], ["hum", 22.43]], [["music", 58.09], ["throbbing", 10.74], ["hum", 6.82]], null, null, null, [["throbbing", 55.86], ["hum", 17.57], ["music", 11.31]], null, [["speech", 42.21], ["music", 17.63], ["throbbing", 12.82]], [["music", 28.99], ["throbbing", 23.65], ["hum", 14.09]], [["throbbing", 24.47], ["hum", 23.09], ["music", 16.9]], [["music", 37.44], ["throbbing", 19.49], ["hum", 12.31]], null], "duration": [47.98, 26.22, 4.62, 1.48, 0.22, 1.8, 8.5, 1.82, 10.29, 3.92, 5.7, 2.85, 1.66]}

annotations_filtered/e80zdJ6pWlI_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 40.68], [44.0, 83.86]], "keep_status": [false, false], "silence_prob": [0.0, 0.0], "audiomae_on_audioset": [null, null], "duration": [35.68, 39.86]}

annotations_filtered/e89qDsf6Q3g_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 8.46], [9.0, 9.95], [24.0, 33.0], [35.0, 35.77], [37.0, 37.12], [38.0, 38.87], [40.0, 40.22], [41.0, 41.72], [42.0, 43.09], [44.0, 45.44], [47.0, 48.47], [52.0, 52.24], [53.0, 59.34], [60.0, 60.71], [62.0, 62.33], [63.0, 64.49], [67.0, 67.9], [80.0, 80.3], [82.0, 82.12], [84.0, 85.06], [86.0, 87.88], [90.0, 90.9], [92.0, 93.85], [96.0, 97.24], [99.0, 100.48], [104.0, 104.04], [108.0, 107.94]], "keep_status": [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], "silence_prob": [46.94, 0.0, 30.86, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 76.86, 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": [[["music", 15.0], ["noise", 12.52], ["synthesizer", 11.84]], null, [["music", 37.67], ["effects unit", 11.98], ["gong", 11.39]], 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": [3.46, 0.95, 9.0, 0.77, 0.12, 0.87, 0.22, 0.72, 1.09, 1.44, 1.47, 0.24, 6.34, 0.71, 0.33, 1.49, 0.9, 0.3, 0.12, 1.06, 1.88, 0.9, 1.85, 1.24, 1.48, 0.04, -0.06]}

annotations_filtered/e8Bn6XVv9ew_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[5.0, 9.86], [15.0, 15.36], [17.0, 18.12], [20.0, 21.39], [23.0, 25.0], [34.0, 36.15], [40.0, 44.49]], "keep_status": [false, false, false, false, false, false, false], "silence_prob": [31.91, 0.0, 0.0, 0.0, 31.11, 30.63, 30.61], "audiomae_on_audioset": [[["fly, housefly", 32.34], ["speech", 29.28], ["insect", 17.25]], null, null, null, [["music", 57.46], ["musical instrument", 6.89], ["guitar", 6.54]], [["music", 68.84], ["musical instrument", 3.91], ["electronic music", 3.75]], [["music", 79.92], ["throbbing", 2.45], ["electronic music", 1.75]]], "duration": [4.86, 0.36, 1.12, 1.39, 2.0, 2.15, 4.49]}

annotations_filtered/e8EAVtsvfxc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[4.0, 5.9], [11.0, 13.63], [15.0, 16.46], [22.0, 24.34], [25.0, 27.01], [28.0, 28.46], [31.0, 31.9], [33.0, 34.05], [35.0, 36.02], [40.0, 39.95], [50.0, 52.08], [56.0, 81.03], [84.0, 85.36], [86.0, 87.35], [89.0, 89.68], [91.0, 91.45], [93.0, 120.28], [121.0, 122.67], [125.0, 127.03], [128.0, 128.95], [131.0, 132.06]], "keep_status": [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, 98.8, 0.0, 99.97, 60.6, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 77.87, 0.0, 0.0, 0.0, 0.0, 98.36, 0.0, 89.01, 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], "duration": [1.9, 2.63, 1.46, 2.34, 2.01, 0.46, 0.9, 1.05, 1.02, -0.05, 2.08, 25.03, 1.36, 1.35, 0.68, 0.45, 27.28, 1.67, 2.03, 0.95, 1.06]}

annotations_filtered/e8HRcYG3Ftg_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[19.0, 20.14], [23.0, 31.01], [33.0, 33.02], [60.0, 63.48], [64.0, 64.91], [66.0, 66.51], [69.0, 69.04], [75.0, 83.81], [85.0, 88.65], [90.0, 93.21], [97.0, 105.14], [106.0, 109.07]], "keep_status": [false, false, false, false, false, false, false, false, false, false, false, false], "silence_prob": [0.0, 33.24, 0.0, 50.16, 0.0, 0.0, 0.0, 75.72, 59.59, 96.77, 89.19, 67.25], "audiomae_on_audioset": [null, [["speech", 54.04], ["snicker", 16.98], ["chuckle, chortle", 5.65]], null, null, null, null, null, null, null, null, null, null], "duration": [1.14, 8.01, 0.02, 3.48, 0.91, 0.51, 0.04, 8.81, 3.65, 3.21, 8.14, 3.07]}

annotations_filtered/e8J56HbIyvc_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[3.0, 3.3], [6.0, 6.56], [12.0, 12.8], [14.0, 14.99], [17.0, 18.64], [20.0, 20.73], [22.0, 22.64], [25.0, 26.2], [32.0, 38.45], [40.0, 41.86], [45.0, 45.54], [48.0, 48.3], [53.0, 53.64], [55.0, 55.83], [59.0, 59.83], [61.0, 61.38], [67.0, 67.91], [69.0, 69.9], [70.0, 71.0], [72.0, 74.17], [76.0, 77.9], [100.0, 102.49], [105.0, 105.19]], "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], "silence_prob": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 81.35, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 100.0, 0.0], "audiomae_on_audioset": [null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null], "duration": [0.3, 0.56, 0.8, 0.99, 1.64, 0.73, 0.64, 1.2, 6.45, 1.86, 0.54, 0.3, 0.64, 0.83, 0.83, 0.38, 0.91, 0.9, 1.0, 2.17, 1.9, 2.49, 0.19]}

annotations_filtered/e8i7WCXqJ94_filtered.json

@@ -0,0 +1 @@
+{"non_speech_segments": [[2.0, 4.28], [5.0, 5.7], [7.0, 7.35], [7.0, 16.28], [18.0, 20.33], [22.0, 22.89], [24.0, 24.7], [31.0, 31.26], [34.0, 34.53], [36.0, 37.52], [42.0, 42.96], [60.0, 61.65]], "keep_status": [true, false, false, false, true, false, false, false, false, false, false, false], "silence_prob": [44.55, 0.0, 0.0, 33.32, 38.43, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], "audiomae_on_audioset": [[["music", 40.76], ["hum", 12.29], ["throbbing", 11.86]], null, null, [["speech", 30.12], ["music", 27.77], ["hum", 13.09]], [["music", 44.26], ["speech", 19.03], ["hum", 3.15]], null, null, null, null, null, null, null], "duration": [2.28, 0.7, 0.35, 9.28, 2.33, 0.89, 0.7, 0.26, 0.53, 1.52, 0.96, 1.65]}