Add read me

README.txt

@@ -0,0 +1,33 @@
+AnuraSet 
+-------------------------------
+Date: 2023-06-15
+-------------------------------
+
+Description
+-------------------------------
+This dataset has 93378 samples of 3s audio which corresponds to 27 hours of human annotations.
+The labels are 42 species of neotropical anurans. In the metadata file, each specie has 
+a column where is 1 if any portion (independete of lenght and quality) are in the sample, and 0 else.
+The dataset was built using soundscape recordings from passive acoustic monitoring in Brazil. The sites are
+INCT17 (42.5%), INCT20955(33%) INCT41 (13.5%) and INCT4 (11%) 
+
+The recordings were preprocessed by resampling the audio to 22050Hz and formatted to 16-bit depth. 
+The recordings are trimmed to a fixed window length, but no filtering was applied.
+The frequency limits are: (1, 10000). 
+The construction used a sliding windows approach of 3s moving 1s (Overlapping of 2/3).
+-------------------------------
+
+Dictionary of data 
+-------------------------------
+-sample_name: the unique identifier of each sample that corresponds to a unique audio file in the audio folder and follows the structure {site}_{date}_{hour}_{start second}_{final second}.wav. The next 5 columns were constructed based on this column. 
+-fname: raw audio filename extracted from a site and used by annotators to create weak labels. 
+-min_t: second where the annotation starts in a fixed window length. 
+-max_t: second where the annotation ends in a fixed window length. 
+-site: identifier of the recording site. 
+-date: datetime of the recording.
+-subset: training or test subset. 
+-species_number: total number of species in each sample. The sum of the next 42 columns per row.
+{species}×42 Binary columns of each species where 1 if some portion of the call is in the sample, 0 else. The 42 species column names are the codes shown in Table 2
+-------------------------------
+
+Contact: jcanas@humboldt.org.co