tyiannak / amvoc

A Python Tool for Analysis of Mouse Vocal Communication

Welcome 👋

I am a Principal Researcher at the Multimedia Analysis Group of the Institute of Informatics and Telecommunication< of the National Center for Scientific Research "Demokritos" and a Machine Learning Director at Behavioral Signals. You can find more info about me at my web page

amvoc's Issues

New UI for comparison between recordings

Basic audio recording functionality

Refactor repository folder structure

Optional

Version issues

Hello, I am having some issues with the current versions of these functions, is there an updated code I could use please? Thank you.

Train a model based on the assigned classes

Read from file simulation (frame-level)

Add the detected cluster to the output csv file

Add the identified cluster per syllable to the offline_filename.csv excel file

Merge very close vocalizations

More clear spectrogram vizualization

Train model(s) based on current vocalizations (multiclass and binary-for-detection)

Cluster to class assignment (by the user)

Deep Spectrogram Feature Extractor

Optimize speed

Add scatter plot in offline UI

Comparison between recordings

UI changes for clustering validation

Refactor dash-related code

Faster (smaller) spectrogram visualization

Add temporal transition plots in offline UI

Issue updating GUI after retraining/annotations

Hello,

I just wanted to share this issue that I came up on here.

I retrained a model using the Dash GUI. After clicking 'Update' it seemed to work fine. After this first re-training I performed about 30 point annotations in one cluster that I wanted to improve, and then I clicked 'Update' again. In terminal it said it was updating, then 'DONE'. Right after this I get the following error:

Traceback (most recent call last):
  File "/Users/cesar/Documents/GitHub/amvoc/src/main.py", line 703, in update_cluster_graph
    if annotation['method'] == method and annotation['number_of_clusters'] == n_clusters and annotation['features_type']==feats_type:
KeyError: 'method'

I couldn't quite figure out how to resolve this.

Another note: I had issues where after performing annotations or retraining I could not save the model again. I resolved this by adding the '.pt' ending to other portions of the code where the model name is needed (two instances).

Online Vocalization Detection

Evaluate Vocalization Detection Vs Offline Mode

Change main.py to also accept supervised vocalization classification (need trained model)

Refactor cluster visualization

Clustering: add features visualization as a subplot next to spectrograms

Clustering: add vocalization spectrograms visualization per cluster as a subplot

Build a production-level auto encoder

Readme Update

Basic V.01 Segmentation Method

The folder app_date is not created automatically the first time the app is run.

First version vocalizations clustering

Improve Unsupervised Vocalization (syllable) detection. Compare to existing outputs (Cezar to send data)

The goal of this issue is to create a robust and more accurate syllable detector (unsupervised).
Towards this end, we will experiment with feature extraction and thresholding parameters. Results will be compared wrt other similar software and/or human generated ground truth.

User can change automatically detected vocalization endpoints