12,571 Hits in 8.1 sec

Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels [article]

Lu Jiang, Di Huang, Mason Liu, Weilong Yang
2020 arXiv   pre-print
Due to the lack of suitable datasets, previous research has only examined deep learning on controlled synthetic label noise, and real-world label noise has never been studied in a controlled setting.  ...  Performing controlled experiments on noisy data is essential in understanding deep learning across noise levels.  ...  LG] 27 Aug 2020 Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels Table 1 . Summary of the difference of images with blue and red noisy labels.  ... 
arXiv:1911.09781v3 fatcat:l2yhsf4oebcs3b7wkxl2uamcfu

AI Online Filters to Real World Image Recognition [article]

Hai Xiao, Jin Shang, Mengyuan Huang
2020 arXiv   pre-print
As demand for intelligent robot control expands to many high level tasks, reinforcement learning and state based models play an increasingly important role.  ...  Deep artificial neural networks, trained with labeled data sets are widely used in numerous vision and robotics applications today.  ...  In this study, we explored premier CV (ISP) image filters (action), AI, Reinforcement Learning (control) and Deep Learning (reward) for a complete end to end learning system.  ... 
arXiv:2002.08242v1 fatcat:ufzdck54fvbslm7s3ab2k2bapq

REAL-M: Towards Speech Separation on Real Mixtures [article]

Cem Subakan, Mirco Ravanelli, Samuele Cornell, François Grondin
2021 arXiv   pre-print
In recent years, deep learning based source separation has achieved impressive results.  ...  Moreover, we observe that the performance trends predicted by our estimator on the REAL-M dataset closely follow those achieved on synthetic benchmarks when evaluating popular speech separation models.  ...  The original utterances are recorded with high-quality microphones in controlled environments where there is no noise and reverberation.  ... 
arXiv:2110.10812v1 fatcat:r44xmbpehbgj3efjzumpljb5ie

Contemplating real-world object classification [article]

Ali Borji
2021 arXiv   pre-print
We also investigate the robustness of models against synthetic image perturbations such as geometric transformations (e.g., scale, rotation, translation), natural image distortions (e.g., impulse noise  ...  Despite this gain, however, we conclude that deep models still suffer drastically on the ObjectNet dataset.  ...  Robustness on real data is a clear challenge for deep neural networks.  ... 
arXiv:2103.05137v2 fatcat:bu2spm6so5dfnlflrozeq6it3a

SuperCaustics: Real-time, open-source simulation of transparent objects for deep learning applications [article]

Mehdi Mousavi, Rolando Estrada
2021 arXiv   pre-print
To address this issue, we present SuperCaustics, a real-time, open-source simulation of transparent objects designed for deep learning applications.  ...  As such, current solutions for this problem employ rigid synthetic datasets, which lack flexibility and lead to severe performance degradation when deployed on real-world scenarios.  ...  In this study, researchers trained deep learning models to estimate depth, surface normals, and occlusion boundaries using a mix of synthetic and real data.  ... 
arXiv:2107.11008v2 fatcat:zuzzrui6enghno72lsiav4i2je

Infant Crying Detection in Real-World Environments [article]

Xuewen Yao, Megan Micheletti, Mckensey Johnson, Edison Thomaz, Kaya de Barbaro
2022 arXiv   pre-print
In this paper, we evaluate several established machine learning approaches including a model leveraging both deep spectrum and acoustic features.  ...  Our findings confirm that a cry detection model trained on in-lab data underperforms when presented with real-world data (in-lab test F1: 0.656, real-world test F1: 0.236), highlighting the value of our  ...  Thus, synthetic datasets that layer additional sounds on clean laboratory datasets are not equivalent to real-world datasets.  ... 
arXiv:2005.07036v6 fatcat:ezqe2ii2ebed5hlgu67cmfcs2i

Real-time Uncertainty Decomposition for Online Learning Control [article]

Jonas Umlauft, Armin Lederer, Thomas Beckers, Sandra Hirche
2020 arXiv   pre-print
We exploit this property in a model-based quadcopter control setting and demonstrate how the controller benefits from a differentiation between aleatoric and epistemic uncertainty in online learning of  ...  Safety-critical decisions based on machine learning models require a clear understanding of the involved uncertainties to avoid hazardous or risky situations.  ...  We evaluate the proposed methods on synthetic and real-world benchmark data sets, and simulate a quadcopter controller, which learns online the disturbances injected by thermals.  ... 
arXiv:2010.02613v2 fatcat:63pljq24nvd47pbeixte7gylb4

Towards efficient models for real-time deep noise suppression [article]

Sebastian Braun, Hannes Gamper, Chandan K.A. Reddy, Ivan Tashev
2021 arXiv   pre-print
With recent research advancements, deep learning models are becoming attractive and powerful choices for speech enhancement in real-time applications.  ...  We show interesting tradeoffs between computational complexity and the achievable speech quality, measured on real recordings using a highly accurate MOS estimator.  ...  EXPERIMENTAL SETUP Dataset We a use large-scale synthetic training set and test on real recordings to ensure generalization of our results to real-world signals.  ... 
arXiv:2101.09249v2 fatcat:e26mygy4vfhc3jznpktdg53f3m

KOVIS: Keypoint-based Visual Servoing with Zero-Shot Sim-to-Real Transfer for Robotics Manipulation [article]

En Yen Puang and Keng Peng Tee and Wei Jing
2020 arXiv   pre-print
Then the visual servoing network learns the motion based on keypoints extracted from the camera image.  ...  We train the deep neural network only in the simulated environment; and the trained model could be directly used for real-world visual servoing tasks. KOVIS consists of two networks.  ...  As a result of recent advancement in deep learning, making use of deep Convolutional Neural Networks (CNN) allows direct learning of the controller for VS, instead of relying on manually handcraft features  ... 
arXiv:2007.13960v1 fatcat:nfyc54hmgvcgxl7bvxeablc2mm

On the Robustness of Intent Classification and Slot Labeling in Goal-oriented Dialog Systems to Real-world Noise [article]

Sailik Sengupta, Jason Krone, Saab Mansour
2021 arXiv   pre-print
By leveraging cross-noise robustness transfer -- training on one noise type to improve robustness on another noise type -- we design aggregate data-augmentation approaches that increase the model performance  ...  across all seven noise types by +10.8% for IC accuracy and +15 points for SL F1 on average.  ...  Synthetic Data Generation Casing Casing variation is common in text modality human-to-bot conversations. Consider for example, the responses "john" vs. "John" vs.  ... 
arXiv:2104.07149v2 fatcat:fj3tvh6b6bgx3bbq5bq4rckzam

Unpaired Point Cloud Completion on Real Scans using Adversarial Training [article]

Xuelin Chen, Baoquan Chen, Niloy J. Mitra
2020 arXiv   pre-print
While these methods have been successfully demonstrated on synthetic data, the approaches cannot be directly used on real scans in absence of suitable paired training data.  ...  As 3D scanning solutions become increasingly popular, several deep learning setups have been developed geared towards that task of scan completion, i.e., plausibly filling in regions there were missed  ...  Note that we do not require the noise characteristics in the two data distributions, i.e., real and synthetic, to be the same.  ... 
arXiv:1904.00069v3 fatcat:d3qrlxopxbckvdexcgobsa5vka

Mask-guided Style Transfer Network for Purifying Real Images [article]

Tongtong Zhao, Yuxiao Yan, Jinjia Peng, Huibing Wang, Xianping Fu
2019 arXiv   pre-print
To solve this problem, the previous method learned a model to improve the realism of the synthetic images.  ...  Different from the previous methods, this paper try to purify real image by extracting discriminative and robust features to convert outdoor real images to indoor synthetic images.  ...  SimGANs [1] refining UnityEyes with its method, try to realistic synthetic image and test on real MPIIGaze dataset.  ... 
arXiv:1903.08152v1 fatcat:7wdcxembxvg4zdbralprqo5hzq

Simulation-based reinforcement learning for real-world autonomous driving [article]

Błażej Osiński, Adam Jakubowski, Piotr Miłoś, Paweł Zięcina, Christopher Galias, Silviu Homoceanu, Henryk Michalewski
2020 arXiv   pre-print
We use reinforcement learning in simulation to obtain a driving system controlling a full-size real-world vehicle.  ...  Based on the extensive evaluation, we analyze how design decisions about perception, control, and training impact the real-world performance.  ...  In some of our experiments, the learned controller outputs the steering command directly.  ... 
arXiv:1911.12905v3 fatcat:c42puc4ydvf6hetv3qcz4cfn4a

Deep-learning Image Reconstruction for Real-time Photoacoustic System [article]

MinWoo Kim, Geng-Shi Jeng, Ivan Pelivanov, Matthew O'Donnell
2020 arXiv   pre-print
Experimental results using synthetic and real datasets confirm that the deep-learning approach provides superior reconstructions compared to conventional methods.  ...  It is designed for real-time clinical applications and trained by large-scale synthetic data mimicking typical microvessel networks.  ...  Here we explore practical PA image reconstruction based on a deep-learning technique suitable for real-time PAUS imaging.  ... 
arXiv:2001.04631v2 fatcat:gmo7nsrllzaqjhldclnlaomfwm

Bridging the Last Mile in Sim-to-Real Robot Perception via Bayesian Active Learning [article]

Jianxiang Feng, Jongseok Lee, Maximilian Durner, Rudolph Triebel
2021 arXiv   pre-print
Therefore, we propose a Sim-to-Real pipeline that relies on deep Bayesian active learning and aims to minimize the manual annotation efforts.  ...  However, when relying only on synthetic data,we encounter the well-known problem of the simulation-to-reality (Sim-to-Real) gap, which is hard to resolve completely in practice.  ...  Then, we rely on deep Bayesian active learning to select the most informative images from a pool of unlabeled real images.  ... 
arXiv:2109.11547v2 fatcat:g3a7rft3xbaspb3kunhcp27y5i
« Previous Showing results 1 — 15 out of 12,571 results