Code of the lileonardo team for the 2021 Emotion and Theme Recognition in Music task of MediaEval 2021

Last update: Aug 02, 2022

Related tags

Overview

Emotion and Theme Recognition in Music

The repository contains code for the submission of the lileonardo team to the 2021 Emotion and Theme Recognition in Music task of MediaEval 2021 (results).

Requirements

python >= 3.7
pip install -r requirements.txt in a virtual environment
Download data from the MTG-Jamendo Dataset in data/jamendo. Audio files go to data/jamendo/mp3 and melspecs to data/jamendo/melspecs.
Process 128 bands mel spectrograms and store them in data/jamendo/melspecs2 by running:
```
python preprocess.py experiments/preprocessing/melspecs2.json
```

Usage

Run python main.py experiments/DIR where DIR contains the parameters.

Parameters are overridable by command line arguments:

python main.py --help

usage: main.py [-h] [--data_dir DATA] [--num_workers NUM] [--restart_training] [--restore_name NAME]
               [--num_epochs EPOCHS] [--learning_rate LR] [--weight_decay WD] [--dropout DROPOUT]
               [--batch_size BS] [--manual_seed SEED] [--model MODEL] [--loss LOSS]
               [--calculate_stats]
               DIRECTORY

Train according to parameters in DIRECTORY

positional arguments:
  DIRECTORY            path of the directory containing parameters

optional arguments:
  -h, --help           show this help message and exit
  --data_dir DATA      path of the directory containing data (default: data)
  --num_workers NUM    number of workers for dataloader (default: 4)
  --restart_training   overwrite previous training (default is to resume previous training)
  --restore_name NAME  name of checkpoint to restore (default: last)
  --num_epochs EPOCHS  override number of epochs in parameters
  --learning_rate LR   override learning rate
  --weight_decay WD    override weight decay
  --dropout DROPOUT    override dropout
  --batch_size BS      override batch size
  --manual_seed SEED   override manual seed
  --model MODEL        override model
  --loss LOSS          override loss
  --calculate_stats    recalculate mean and std of data (default is to calculate only when they
                       don't exist in parameters)

Ensemble predictions

The predictions are averaged by running:

python average.py --outputs experiments/convs-m96*/predictions/test-last-swa-outputs.npy --targets experiments/convs-m96*/predictions/test-last-swa-targets.npy --preds_path predictions/convs.npy

python average.py --outputs experiments/filters-m128*/predictions/test-last-swa-outputs.npy --targets experiments/filters-m128*/predictions/test-last-swa-targets.npy --preds_path predictions/filters.npy

python average.py --outputs predictions/convs.npy predictions/filters.npy --targets predictions/targets.npy

Code of the lileonardo team for the 2021 Emotion and Theme Recognition in Music task of MediaEval 2021

Related tags

Overview

Emotion and Theme Recognition in Music

Requirements

Usage

Ensemble predictions

Owner

Vincent Bour

ARKitScenes - A Diverse Real-World Dataset for 3D Indoor Scene Understanding Using Mobile RGB-D Data

Proposed n-stage Latent Dirichlet Allocation method - A Novel Approach for LDA

RRL: Resnet as representation for Reinforcement Learning

Hands-On Machine Learning for Algorithmic Trading, published by Packt

Deeprl - Standard DQN and dueling network for simple games

Magisk module to enable hidden features on Android 12 Developer Preview 1.

A list of Machine Learning Art Colabs

This project aims to segment 4 common retinal lesions from Fundus Images.

Code for ACM MM 2020 paper "NOH-NMS: Improving Pedestrian Detection by Nearby Objects Hallucination"

Text and code for the forthcoming second edition of Think Bayes, by Allen Downey.

PyTorch implementation for "HyperSPNs: Compact and Expressive Probabilistic Circuits", NeurIPS 2021

Official implementation for (Refine Myself by Teaching Myself : Feature Refinement via Self-Knowledge Distillation, CVPR-2021)

deep-table implements various state-of-the-art deep learning and self-supervised learning algorithms for tabular data using PyTorch.

This is the repository for the NeurIPS-21 paper [Contrastive Graph Poisson Networks: Semi-Supervised Learning with Extremely Limited Labels].

Research shows Google collects 20x more data from Android than Apple collects from iOS. Block this non-consensual telemetry using pihole blocklists.

Code to reproduce the experiments from our NeurIPS 2021 paper " The Limitations of Large Width in Neural Networks: A Deep Gaussian Process Perspective"

ZSL-KG is a general-purpose zero-shot learning framework with a novel transformer graph convolutional network (TrGCN) to learn class representation from common sense knowledge graphs.

Data, model training, and evaluation code for "PubTables-1M: Towards a universal dataset and metrics for training and evaluating table extraction models".

This repository is the official implementation of Unleashing the Power of Contrastive Self-Supervised Visual Models via Contrast-Regularized Fine-Tuning (NeurIPS21).

Script that receives an Image (original) and a set of images to be used as "pixels" in reconstruction of the Original image using the set of images as "pixels"