1. |
- Andersson, Matilda, et al.
(författare)
-
Augmentation Strategies for Self-Supervised Representation Learning from Electrocardiograms
- 2023
-
Ingår i: 31st European Signal Processing Conference, EUSIPCO 2023 - Proceedings. - 2219-5491. - 9789464593600 ; , s. 1075-1079
-
Konferensbidrag (refereegranskat)abstract
- In this paper, we investigate the effects of different augmentation strategies in self-supervised representation learning from electrocardiograms. Our study examines the impact of random resized crop and time out on downstream performance. We also consider the importance of the signal length. Furthermore, instead of using two augmented copies of the sample as a positive pair, we suggest augmenting only one. The second signal is kept as the original signal. These different augmentation strategies are investigated in the context of pre-training and fine-tuning, following the different self-supervised learning frameworks BYOL, SimCLR, and VICReg. We formulate the downstream task as a multi-label classification task using a public dataset containing ECG recordings and annotations. In our experiments, we demonstrate that self-supervised learning can consistently outperform classical supervised learning when configured correctly. These findings are of particular importance in the medical domain, as the medical labeling process is particularly expensive, and clinical ground truth is often difficult to define. We are hopeful that our findings will be a catalyst for further research into augmentation strategies in self-supervised learning to improve performance in the detection of cardiovascular disease.
|
|
2. |
- Aström, Kalle, et al.
(författare)
-
Extension of Time-Difference-of-Arrival Self Calibration Solutions Using Robust Multilateration
- 2021
-
Ingår i: 29th European Signal Processing Conference, EUSIPCO 2021 - Proceedings. - 2076-1465 .- 2219-5491. - 9789082797060 ; , s. 870-874
-
Konferensbidrag (refereegranskat)abstract
- Recent advances in robust self-calibration have made it possible to estimate microphone positions and at least partial sound source positions using ambient sound. However, there are limits on how well sound source paths can be recovered using state-of-the-art techniques. In this paper we develop and evaluate several techniques to extend partial and incomplete solutions. We present minimal solvers for sound source positioning using non-overlapping pairs of microphone positions and their respective time-difference measurements, and show how these new solvers can be used in a hypothesis and test setting. We also investigate techniques that exploit temporal smoothness of the sound source paths. The different techniques are evaluated on both real and synthetic data, and compared to several state-of-the-art techniques for time-difference-of-arrival multilateration.
|
|
3. |
- Flood, Gabrielle, et al.
(författare)
-
Efficient Merging of Maps and Detection of Changes
- 2019
-
Ingår i: Image Analysis - 21st Scandinavian Conference, SCIA 2019, Proceedings. - Cham : Springer International Publishing. - 0302-9743 .- 1611-3349. - 9783030202040 ; 11482 LNCS, s. 348-360
-
Konferensbidrag (refereegranskat)abstract
- With the advent of cheap sensors and computing capabilities as well as better algorithms it is now possible to do structure from motion using crowd sourced data. Individual estimates of a map can be obtained using structure from motion (SfM) or simultaneous localization and mapping (SLAM) using e.g. images, sound or radio. However the problem of map merging as used for collaborative SLAM needs further attention. In this paper we study the basic principles behind map merging and collaborative SLAM. We develop a method for merging maps – based on a small memory footprint representation of individual maps – in a way that is computationally efficient. We also demonstrate how the same framework can be used to detect changes in the map. This makes it possible to remove inconsistent parts before merging the maps. The methods are tested on both simulated and real data, using both sensor data from radio sensors and from cameras.
|
|
4. |
- Flood, Gabrielle, et al.
(författare)
-
Estimating Uncertainty in Time-difference and Doppler Estimates
- 2018
-
Ingår i: ICPRAM 2018 - Proceedings of the 7th International Conference on Pattern Recognition Applications and Methods - Volume 1: ICPRAM. - : SCITEPRESS - Science and Technology Publications. - 9789897582769 ; , s. 245-253
-
Konferensbidrag (refereegranskat)abstract
- Sound and radio can be used to estimate the distance between a transmitter and a sender by correlating the emitted and received signal. Alternatively by correlating two received signals it is possible to estimate distance difference. Such methods can be divided into methods that are robust to noise and reverberation, but give limited precision and sub-sample refinements that are sensitive to noise, but give higher precision when initialized close to the real translation. In this paper we develop stochastic models that can explain the limits in the precision of such sub-sample time-difference estimates. Using such models we provide new methods for precise estimates of time-differences as well as Doppler effects. The method is verified on both synthetic and real data.
|
|
5. |
|
|
6. |
- Flood, Gabrielle
(författare)
-
Mapping and Merging Using Sound and Vision : Automatic Calibration and Map Fusion with Statistical Deformations
- 2021
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Over the last couple of years both cameras, audio and radio sensors have become cheaper and more common in our everyday lives. Such sensors can be used to create maps of where the sensors are positioned and the appearance of the surroundings. For sound and radio, the process of estimating the sender and receiver positions from time of arrival (TOA) or time-difference of arrival (TDOA) measurements is referred to as automatic calibration. The corresponding process for images is to estimate the camera positions as well as the positions of the objects captured in the images. This is called structure from motion (SfM) or visual simultaneous localisation and mapping (SLAM). In this thesis we present studies on how to create such maps, divided into three parts: to find accurate measurements; robust mapping; and merging of maps.The first part is treated in Paper I and involves finding precise – on a subsample level – TDOA measurements. These types of subsample refinements give a high precision, but are sensitive to noise. We present an explicit expression for the variance of the TDOA estimate and study the impact that noise in the signals has. Exact measurements is an important foundation for creating accurate maps. The second part of this thesis includes Papers II–V and covers the topic of robust self-calibration using one-dimensional signals, such as sound or radio. We estimate both sender and receiver positions using TOA and TDOA measurements. The estimation process is divided in two parts, where the first is specific for TOA or TDOA and involves solving a relaxed version of the problem. The second step is common for different types of problems and involves an upgrade from the relaxed solution to the sought parameters. In this thesis we present numerically stable minimal solvers for both these steps for some different setups with senders and receivers. We also suggest frameworks for how to use these solvers together with RANSAC to achieve systems that are robust to outliers, noise and missing data. Additionally, in the last paper we focus on extending self-calibration results, especially for the sound source path, which often cannot be fully reconstructed immediately. The third part of the thesis, Papers VI–VIII, is concerned with the merging of already estimated maps. We mainly focus on maps created from image data, but the methods are applicable to sparse 3D maps coming from different sensor modalities. Merging of maps can be advantageous if there are several map representations of the same environment, or if there is a need for adding new information to an already existing map. We suggest a compact map representation with a small memory footprint, which we then use to fuse maps efficiently. We suggest one method for fusion of maps that are pre-aligned, and one where we additionally estimate the coordinate system. The merging utilises a compact approximation of the residuals and allows for deformations in the original maps. Furthermore, we present minimal solvers for 3D point matching with statistical deformations – which increases the number of inliers when the original maps contain errors.
|
|
7. |
- Flood, Gabrielle, et al.
(författare)
-
Minimal Solvers for Point Cloud Matching with Statistical Deformations
- 2022
-
Ingår i: 2022 26th International Conference on Pattern Recognition (ICPR). - 9781665490634
-
Konferensbidrag (refereegranskat)abstract
- An important issue in simultaneous localisation and mapping is how to match and merge individual local maps into one global map. This is addressed within the field of robotics and is crucial for multi-robot SLAM. There are a number of different ways to solve this task depending on the representation of the map. To take advantage of matching and merging methods that allow for deformations of the local maps it is important to find feature matches that capture such deformations. In this paper we present minimal solvers for point cloud matching using statistical deformations. The solvers use either three or four point matches. These solve for either rigid or similarity transformation as well as shape deformation in the direction of the most important modes of variation. Given an initial set of tentative matches based on, for example, feature descriptors or machine learning we use these solvers in a RANSAC loop to remove outliers among the tentative matches. We evaluate the methods on both synthetic and real data and compare them to RANSAC methods based on Procrustes and demonstrate that the proposed methods improve on the current state-of-the-art.
|
|
8. |
- Flood, Gabrielle, et al.
(författare)
-
Stochastic Analysis of Time-Difference and Doppler Estimates for Audio Signals
- 2019
-
Ingår i: Pattern Recognition Applications and Methods - 7th International Conference, ICPRAM 2018, Revised Selected Papers. - Cham : Springer International Publishing. - 1611-3349 .- 0302-9743. - 9783030054984 ; 11351 LNCS, s. 116-138
-
Konferensbidrag (refereegranskat)abstract
- Pairwise comparison of sound and radio signals can be used to estimate the distance between two units that send and receive signals. In a similar way it is possible to estimate differences of distances by correlating two received signals. There are essentially two groups of such methods, namely methods that are robust to noise and reverberation, but give limited precision and sub-sample refinements that are more sensitive to noise, but also give higher precision when they are initialized close to the real translation. In this paper, we present stochastic models that can explain the precision limits of such sub-sample time-difference estimates. Using these models new methods are provided for precise estimates of time-differences as well as Doppler effects. The developed methods are evaluated and verified on both synthetic and real data.
|
|
9. |
- Flood, Gabrielle
(författare)
-
Towards Precise Localisation : Subsample Methods, Efficient Estimation and Merging of Maps
- 2019
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Over the last couple of years audio and radio sensors have become cheaper and more common in our everyday life. Such sensors can be used to form a network, from which one can obtain distance measures by correlating the different received signals. One example of such distance measures is time-difference of arrival measurements (TDoA), which can be used to estimate the positions of the senders and receivers. The result is a 3D map of the environment, similar to what you get from doing structure from motion (SfM) with images. If a new sensor appears, the map can in turn be used to determine the position of that sensor, i.e. for localisation. In this thesis we present three studies that take us towards precise localisation. Paper I involves finding exact — on a subsample level — TDoA measurements. These types of subsample refinements give a higher precision, but are sensitive to noise. We present an explicit expression for the variance of the TDoA estimate and study the impact that noise in the signals have. In Paper III TDoA measurements are used to estimate sender and receiver positions in an efficient way. We present a new initialisation approach followed by a scheme for performing local optimisation for TDoA data with constant offset, i.e. when the sound events are repetitive with some constant period. The sender and receiver positions together constitute a map of the environment and such maps are studied in Paper II. Assuming that we have a number of different map representations of the same environment — coming from either sound, radio or image data — we present an algorithm for how to merge these representations into one map, in an efficient way using only a small memory footprint representation. The final map has a higher precision and the method can also be used to detect changes that have occurred between the creation of the different map representations. Thus, altogether, we present a number of improvements of the localisation process. We perform analysis as well as experimental evaluation of each of these improvements.
|
|
10. |
- Gillsjö, David, et al.
(författare)
-
Polygon Detection for Room Layout Estimation using Heterogeneous Graphs and Wireframes
- 2023
-
Ingår i: Proceedings - 2023 IEEE/CVF International Conference on Computer Vision Workshops, ICCVW 2023. - 9798350307443 ; , s. 1-10
-
Konferensbidrag (refereegranskat)abstract
- This paper presents a neural network based semantic plane detection method utilizing polygon representations. The method can for example be used to solve room layout estimations tasks and is built on, combines and further develops several different modules from previous research. The network takes an RGB image and estimates a wireframe as well as a feature space using an hourglass backbone. From these, line and junction features are sampled. The lines and junctions are then represented as an undirected graph, from which polygon representations of the sought planes are obtained. Two different methods for this last step are investigated, where the most promising method is built on a heterogeneous graph transformer. The final output is in all cases a projection of the semantic planes in 2D. The methods are evaluated on the Structured3D dataset and we investigate the performance both using sampled and estimated wireframes. The experiments show the potential of the graph-based method by outperforming state of the art methods in Room Layout estimation in the 2D metrics using synthetic wireframe detections.
|
|