Structural Segmentation with Convolutional Neural Networks Mirex Submission
نویسندگان
چکیده
This submission to the MIREX 2015 Music Structural Segmentation task employs a Convolutional Neural Network (CNN) to identify boundaries within a piece of digital audio. The network was trained on a combination of mel-scaled log-magnitude spectrograms (MLSs) and selfsimilarity lag matrices (SSLMs) with two-level human structural annotations following the SALAMI guidelines. It is based on our work presented in Grill and Schlüter [3]. Apart from detecting boundaries, our submission also attempts to assign labels to the resulting segments using a simple model based on 2D-DCTs and a cosine distance measure.
منابع مشابه
A hierarchical Convolutional Neural Network for Segmentation of Stroke Lesion in 3D Brain MRI
Introduction: Brain tumors such as glioma are among the most aggressive lesions, which result in a very short life expectancy in patients. Image segmentation is highly essential in medical image analysis with applications, particularly in clinical practices to treat brain tumors. Accurate segmentation of magnetic resonance data is crucial for diagnostic purposes, planning surgical treatments, a...
متن کاملA hierarchical Convolutional Neural Network for Segmentation of Stroke Lesion in 3D Brain MRI
Introduction: Brain tumors such as glioma are among the most aggressive lesions, which result in a very short life expectancy in patients. Image segmentation is highly essential in medical image analysis with applications, particularly in clinical practices to treat brain tumors. Accurate segmentation of magnetic resonance data is crucial for diagnostic purposes, planning surgical treatments, a...
متن کاملA multi-scale convolutional neural network for automatic cloud and cloud shadow detection from Gaofen-1 images
The reconstruction of the information contaminated by cloud and cloud shadow is an important step in pre-processing of high-resolution satellite images. The cloud and cloud shadow automatic segmentation could be the first step in the process of reconstructing the information contaminated by cloud and cloud shadow. This stage is a remarkable challenge due to the relatively inefficient performanc...
متن کاملParallel Convolutional Neural Networks for Music Genre and Mood Classification
Our approach to the MIREX 2016 Train/Test Classification Tasks for Genre, Mood and Composer detection is based on an approach combining Melspectrogram transformed audio and Convolutional Neural Networks (CNN). We utilize two different CNN architectures, a sequential one, and a parallel one, the latter aiming at capturing both temporal and timbral information in two different pipelines, which ar...
متن کاملEstimation of Hand Skeletal Postures by Using Deep Convolutional Neural Networks
Hand posture estimation attracts researchers because of its many applications. Hand posture recognition systems simulate the hand postures by using mathematical algorithms. Convolutional neural networks have provided the best results in the hand posture recognition so far. In this paper, we propose a new method to estimate the hand skeletal posture by using deep convolutional neural networks. T...
متن کامل