diff --git a/README_get_time_freq_detection.md b/README_get_time_freq_detection.md
new file mode 100644
index 0000000000000000000000000000000000000000..12e3d350a9cb4fdb0c3404f1c23a49cff40f21f6
--- /dev/null
+++ b/README_get_time_freq_detection.md
@@ -0,0 +1,13 @@
+
+
+## Output (file.csv) of get_time_freq_detection.py
+
+| **_COLUMN NAME_** | **espece**   | **conf**     | **annot**    | **midl**     | **freq_center** | **freq_min** | **freq_max** | **start**    | **stop**     | **duration**    | **station**  | **site**     | **date**  | **date_t** | **normalized_hour** |
+| ----------------: | -----------: | -----------: | -----------: | -----------: | --------------: | -----------: | -----------: | -----------: | -----------: | --------------: | -----------: | -----------: | --------: | ---------: | ------------------: |
+| **_CLASS_**       | vocalization | vocalization | vocalization | vocalization | vocalization    | vocalization | vocalization | file         | file         | file            | file         | file         | file      | file       | file                |
+| **_DESCRIPTION_** |label number ( ex : from 0 to 32  for 32 species labels)       | confidence level of the prediction [value : 0-1] the higher it gets, the more confident the network get      | name of species corresponding to « espece » column (ex : espece =0, annot = wtsp)      | time when the vocalize is detected :  (stop - start)/2      | centroid frequency of the detection (because the detection is inside a boundingbox : freq_center=height of the boundingbox/2)      | minimum frequency of the boundingbox          | maximal frequency of the vocalize          | filename with the vocalize detected          | filename with the vocalize detected          | path to the file inside the server          | sample frequency of the file with the vocalize          | duration of the file with the vocalize (seconds)          | station of the recordings          | site name of the recordings          | date of the recording with an hour precision          | date of the recording without taking into acompt the hours          | number of  detections/hour of recording (ex: file duration=30min, normalized_hour=2)          |
+
+    
+
+
+