Hugging Face's logo

Datasets:
zed-m97
/
nano4m-Audio

Tasks:
Audio Classification
Text-to-Audio
Audio-to-Audio
Modalities:
Text
Audio
Languages:
English
Size:
1K - 10K
ArXiv:
Tags:
audio
multimodal
nano4m
4m
vggsound
encodec
License:
Dataset card Data Studio Files
xet
 Community
1
zed-m97 commited on
Commit
c698b8f
·
verified ·
1 Parent(s): e0cb742

Add 12-class team stats

Browse files
Files changed (1) hide show
  1. stats_team.md +195 -0
stats_team.md ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Person-2 dataset stats
2
+
3
+ Manifest: `data/manifest_team.csv` (6529 rows)
4
+
5
+ ## Status breakdown
6
+
7
+ | status | count | % of total |
8
+ |---|---:|---:|
9
+ | `OK` | 4905 | 75.1% |
10
+ | `FAIL_DL` | 1203 | 18.4% |
11
+ | `MISSING` | 372 | 5.7% |
12
+ | `FAIL_CLIP` | 27 | 0.4% |
13
+ | `FAIL_CLAP` | 20 | 0.3% |
14
+ | `FAIL_PEAKS` | 1 | 0.0% |
15
+ | `FAIL_CUT` | 1 | 0.0% |
16
+
17
+ ## Per-class survival (OK / total)
18
+
19
+ | class | train OK | test OK | total OK | total raw | survival |
20
+ |---|---:|---:|---:|---:|---:|
21
+ | cat meowing | 449 | 41 | 490 | 588 | 83.3% |
22
+ | chicken clucking | 454 | 37 | 491 | 625 | 78.6% |
23
+ | cow lowing | 98 | 13 | 111 | 199 | 55.8% |
24
+ | coyote howling | 344 | 41 | 385 | 490 | 78.6% |
25
+ | dog barking | 599 | 38 | 637 | 767 | 83.1% |
26
+ | donkey, ass braying | 302 | 33 | 335 | 438 | 76.5% |
27
+ | duck quacking | 501 | 38 | 539 | 602 | 89.5% |
28
+ | elephant trumpeting | 183 | 34 | 217 | 310 | 70.0% |
29
+ | horse neighing | 325 | 13 | 338 | 587 | 57.6% |
30
+ | lions roaring | 639 | 4 | 643 | 964 | 66.7% |
31
+ | pig oinking | 264 | 12 | 276 | 444 | 62.2% |
32
+ | sheep bleating | 405 | 38 | 443 | 515 | 86.0% |
33
+
34
+ ## Failure-mode breakdown per class
35
+
36
+ | class | FAIL_DL | FAIL_CLIP | FAIL_CLAP | FAIL_PEAKS | MISSING |
37
+ |---|---:|---:|---:|---:|---:|
38
+ | cat meowing | 88 | 6 | 4 | 0 | 0 |
39
+ | chicken clucking | 0 | 0 | 0 | 0 | 134 |
40
+ | cow lowing | 87 | 1 | 0 | 0 | 0 |
41
+ | coyote howling | 90 | 9 | 5 | 1 | 0 |
42
+ | dog barking | 127 | 3 | 0 | 0 | 0 |
43
+ | donkey, ass braying | 0 | 0 | 0 | 0 | 103 |
44
+ | duck quacking | 0 | 0 | 0 | 0 | 63 |
45
+ | elephant trumpeting | 80 | 8 | 5 | 0 | 0 |
46
+ | horse neighing | 245 | 0 | 4 | 0 | 0 |
47
+ | lions roaring | 318 | 0 | 2 | 0 | 0 |
48
+ | pig oinking | 168 | 0 | 0 | 0 | 0 |
49
+ | sheep bleating | 0 | 0 | 0 | 0 | 72 |
50
+
51
+ ## CLIP score distribution (OK only)
52
+
53
+ Non-NaN OK rows: 1720 / 4905 (NaN = smoke-test residuals predating canonical log)
54
+
55
+ threshold = 0.2, mean = 0.283, std = 0.025, min = 0.200, max = 0.366
56
+
57
+ ```
58
+ [0.200, 0.211) 20 ##
59
+ [0.211, 0.222) 18 ##
60
+ [0.222, 0.233) 38 ####
61
+ [0.233, 0.244) 45 #####
62
+ [0.244, 0.255) 82 #########
63
+ [0.255, 0.266) 159 ##################
64
+ [0.266, 0.277) 283 #################################
65
+ [0.277, 0.289) 338 #######################################
66
+ [0.289, 0.300) 346 ########################################
67
+ [0.300, 0.311) 204 ########################
68
+ [0.311, 0.322) 122 ##############
69
+ [0.322, 0.333) 26 ###
70
+ [0.333, 0.344) 25 ###
71
+ [0.344, 0.355) 11 #
72
+ [0.355, 0.366) 3
73
+ ```
74
+
75
+ Per-class CLIP mean ± std:
76
+
77
+ | class | mean | std | min | max |
78
+ |---|---:|---:|---:|---:|
79
+ | cat meowing | 0.275 | 0.020 | 0.203 | 0.329 |
80
+ | chicken clucking | nan | nan | nan | nan |
81
+ | cow lowing | nan | nan | nan | nan |
82
+ | coyote howling | 0.284 | 0.036 | 0.200 | 0.366 |
83
+ | dog barking | 0.288 | 0.018 | 0.201 | 0.347 |
84
+ | donkey, ass braying | nan | nan | nan | nan |
85
+ | duck quacking | nan | nan | nan | nan |
86
+ | elephant trumpeting | 0.284 | 0.024 | 0.202 | 0.334 |
87
+ | horse neighing | nan | nan | nan | nan |
88
+ | lions roaring | nan | nan | nan | nan |
89
+ | pig oinking | nan | nan | nan | nan |
90
+ | sheep bleating | nan | nan | nan | nan |
91
+
92
+ ## CLAP score distribution (OK only, raw logits)
93
+
94
+ Non-NaN OK rows: 1720 / 4905
95
+
96
+ threshold = 0.2, mean = 11.71, std = 3.02, min = 0.28, max = 20.18
97
+
98
+ ```
99
+ [ 0.28, 1.60) 16 ##
100
+ [ 1.60, 2.93) 13 #
101
+ [ 2.93, 4.26) 27 ###
102
+ [ 4.26, 5.58) 36 ###
103
+ [ 5.58, 6.91) 42 ####
104
+ [ 6.91, 8.24) 64 ######
105
+ [ 8.24, 9.56) 105 ##########
106
+ [ 9.56, 10.89) 218 ####################
107
+ [ 10.89, 12.22) 347 #################################
108
+ [ 12.22, 13.54) 426 ########################################
109
+ [ 13.54, 14.87) 235 ######################
110
+ [ 14.87, 16.20) 122 ###########
111
+ [ 16.20, 17.52) 53 #####
112
+ [ 17.52, 18.85) 15 #
113
+ [ 18.85, 20.18) 1
114
+ ```
115
+
116
+ Per-class CLAP mean ± std:
117
+
118
+ | class | mean | std | min | max |
119
+ |---|---:|---:|---:|---:|
120
+ | cat meowing | 12.03 | 2.83 | 0.72 | 18.77 |
121
+ | chicken clucking | nan | nan | nan | nan |
122
+ | cow lowing | nan | nan | nan | nan |
123
+ | coyote howling | 10.63 | 3.30 | 0.28 | 17.08 |
124
+ | dog barking | 12.48 | 2.01 | 3.36 | 18.70 |
125
+ | donkey, ass braying | nan | nan | nan | nan |
126
+ | duck quacking | nan | nan | nan | nan |
127
+ | elephant trumpeting | 10.64 | 4.32 | 0.52 | 20.18 |
128
+ | horse neighing | nan | nan | nan | nan |
129
+ | lions roaring | nan | nan | nan | nan |
130
+ | pig oinking | nan | nan | nan | nan |
131
+ | sheep bleating | nan | nan | nan | nan |
132
+
133
+ ## Audio sample-count uniformity
134
+
135
+ unique sample counts: 1 (target = 1, all clips at 41,040 = 1.7100 s @ 24 kHz)
136
+ 0 samples (0.0000 s): 4905 clips
137
+
138
+ ## Qualitative samples (3 best + 3 worst CLIP score per class)
139
+
140
+ Copied to `data/_qualitative/{class}/{best,worst}_clip/{stem}.{wav,jpg}` for eyeball/listen.
141
+
142
+ ### cat meowing
143
+
144
+ | rank | stem | split | clip | clap |
145
+ |---|---|---|---:|---:|
146
+ | best #1 | `eo2HskB5Y8I_000015` | train | 0.329 | 10.11 |
147
+ | best #2 | `rjm_phhaTqg_000012` | train | 0.325 | 15.27 |
148
+ | best #3 | `84joiXni4uk_000030` | train | 0.321 | 13.49 |
149
+ | worst #1 | `18HRNQyo9mg_000025` | test | 0.203 | 6.14 |
150
+ | worst #2 | `_V24O52TldA_000004` | train | 0.204 | 14.73 |
151
+ | worst #3 | `1b1b-9_AcFg_000030` | train | 0.205 | 14.53 |
152
+
153
+ ### coyote howling
154
+
155
+ | rank | stem | split | clip | clap |
156
+ |---|---|---|---:|---:|
157
+ | best #1 | `1J5LA2V-68k_000011` | train | 0.366 | 8.46 |
158
+ | best #2 | `1J5LA2V-68k_000043` | train | 0.366 | 11.50 |
159
+ | best #3 | `XA6RBioito4_000067` | test | 0.364 | 9.92 |
160
+ | worst #1 | `1Jb2bUPCU8U_000056` | train | 0.200 | 13.40 |
161
+ | worst #2 | `hRphojIYqiw_000018` | train | 0.201 | 8.60 |
162
+ | worst #3 | `WKbADRKJLo0_000138` | test | 0.205 | 9.26 |
163
+
164
+ ### dog barking
165
+
166
+ | rank | stem | split | clip | clap |
167
+ |---|---|---|---:|---:|
168
+ | best #1 | `urmwibHFOHc_000000` | train | 0.347 | 12.95 |
169
+ | best #2 | `wqMoQkNqb3U_000000` | train | 0.323 | 9.65 |
170
+ | best #3 | `VxsdrQ7a5eg_000000` | train | 0.322 | 11.61 |
171
+ | worst #1 | `UNzasfgSrTo_000010` | train | 0.201 | 12.46 |
172
+ | worst #2 | `LHjIPTqS94k_000030` | train | 0.206 | 18.42 |
173
+ | worst #3 | `m98BKPAD-84_000020` | test | 0.217 | 13.39 |
174
+
175
+ ### elephant trumpeting
176
+
177
+ | rank | stem | split | clip | clap |
178
+ |---|---|---|---:|---:|
179
+ | best #1 | `tY0ba7snoek_000005` | train | 0.334 | 16.31 |
180
+ | best #2 | `fy3-UPb_z4k_000017` | train | 0.332 | 15.09 |
181
+ | best #3 | `itTQe0xQUyw_000019` | train | 0.329 | 13.86 |
182
+ | worst #1 | `r-larRRhtco_000000` | train | 0.202 | 15.77 |
183
+ | worst #2 | `r-larRRhtco_000011` | train | 0.204 | 15.60 |
184
+ | worst #3 | `lEDHRh8gfm8_000094` | train | 0.205 | 3.16 |
185
+
186
+ ## Modality presence across OK rows
187
+
188
+ | modality | present / total | % |
189
+ |---|---:|---:|
190
+ | raw audio (wav) | 2261 / 4905 | 46.1% |
191
+ | raw image (jpg) | 2261 / 4905 | 46.1% |
192
+ | EnCodec tokens (tok_audio@256) | 4884 / 4905 | 99.6% |
193
+ | scene_desc JSON | 4884 / 4905 | 99.6% |
194
+ | Cosmos RGB tokens (tok_rgb@256, GPU step) | 4884 / 4905 | 99.6% |
195
+