40,157 Hits in 8.4 sec

Machine learning and deep learning

Christian Janiesch, Patrick Zschech, Kai Heinrich
2021 Electronic Markets  
Machine learning describes the capacity of systems to learn from problem-specific training data to automate the process of analytical model building and solve associated tasks.  ...  In this article, we summarize the fundamentals of machine learning and deep learning to generate a broader understanding of the methodical underpinning of current intelligent systems.  ...  The concept of transfer learning allows models that are trained on general datasets (e.g., large-scale image datasets) to be specialized for specific tasks by using a considerably smaller dataset that  ... 
doi:10.1007/s12525-021-00475-2 fatcat:k6mhktpp3jdy7jgwoipznzks6e

Multi-View Priors for Learning Detectors from Sparse Viewpoint Data [article]

Bojan Pepik, Michael Stark, Peter Gehler, Bernt Schiele
2014 arXiv   pre-print
be used to facilitate the learning of a detector for a target class.  ...  In this paper, we address this issue from the perspective of transfer learning, and design an object class model that explicitly leverages correlations between visual features.  ...  Multi-view transfer learning We consider the scenario of transfer learning for object models.  ... 
arXiv:1312.6095v2 fatcat:2clhrtpzwvfyhdwruzihhqnspa

Unsupervised Domain Adaptation for Object Detection via Cross-Domain Semi-Supervised Learning [article]

Fuxun Yu, Di Wang, Yinpeng Chen, Nikolaos Karianakis, Tong Shen, Pei Yu, Dimitrios Lymberopoulos, Sidi Lu, Weisong Shi, Xiang Chen
2021 arXiv   pre-print
However, without ground truth labels, most prior works on UDA for object detection tasks can only perform coarse image-level and/or feature-level adaptation by using adversarial learning methods.  ...  To enable SSL for cross-domain object detection, we propose fine-grained domain transfer, progressive-confidence-based label sharpening and imbalanced sampling strategy to address two challenges: (i) non-identical  ...  These models are then used to translate the source domain data for training object detection model using Faster-RCNN.  ... 
arXiv:1911.07158v5 fatcat:avo3zydua5dalo7e6ggnik3wuy

Prescriptive and Descriptive Approaches to Machine-Learning Transparency [article]

David Adkins, Bilal Alsallakh, Adeel Cheema, Narine Kokhlikyan, Emily McReynolds, Pushkar Mishra, Chavez Procope, Jeremy Sawruk, Erin Wang, Polina Zvyagina
2022 arXiv   pre-print
We showcase our proposal with an example in small object detection, and demonstrate how Method Cards can communicate key considerations for model developers.  ...  Specialized documentation techniques have been developed to communicate key facts about machine-learning (ML) systems and the datasets and models they rely on.  ...  [2] , leading to skewness in the learned filters. • The default 0-padding used in ResNet was shown to impact small object detection [3] , leading to blind spots. • Transfer learning from a supervised  ... 
arXiv:2204.13582v1 fatcat:toluuiketzcttiiuljkq6mkjfq

Crowd Sourcing based Active Learning Approach for Parking Sign Recognition [article]

Humayun Irshad, Qazaleh Mirsharif, Jennifer Prendki
2018 arXiv   pre-print
Active learning techniques are being progressively adopted to accelerate the development of machine learning solutions by allowing the model to query the data they learn from.  ...  Deep learning models have been used extensively to solve real-world problems in recent years. The performance of such models relies heavily on large amounts of labeled data for training.  ...  Object Detection Models Deep learning-based object detection models have shown great performance in object detection and generalize very well (Huang et al. 2017 ).  ... 
arXiv:1812.01081v1 fatcat:r4wlomq2bzea5fuwb6b7bfyszy

Learn from Object Counting: Crowd Counting with Meta‐learning

Changtong Zan, Baodi Liu, Weili Guan, Kai Zhang, Weifeng Liu
2021 IET Image Processing  
The objective of crowd counting is to learn a counter that can estimate the number of people in a single image.  ...  To address such a situation, utilizing object counting data in few-shot scenes is considered and an efficient algorithm to extract the meta-information is proposed, thus improving the accuracy and convergence  ...  The meta-learning method Meta-learning, namely learning-to-learn, attempts to extract meta-information from the meta-training process, which can help the model converge faster and better in training new  ... 
doi:10.1049/ipr2.12241 fatcat:7o4kpwrwuvgt7cqsdvjwta74ii

Cap2Det: Learning to Amplify Weak Caption Supervision for Object Detection [article]

Keren Ye, Mingda Zhang, Adriana Kovashka, Wei Li, Danfeng Qin, Jesse Berent
2019 arXiv   pre-print
Our discovery provides an opportunity for learning detection models from noisy but more abundant and freely-available caption data.  ...  However, straightforward approaches to using such data for WSOD wastefully discard captions that do not exactly match object names.  ...  Then, it is used to guide the learning of theŝ (k+1) using Eq.5 in the paper.  ... 
arXiv:1907.10164v3 fatcat:qbuq2q3oibbhrbeza6lwzzdppe

Shuffle-Then-Assemble: Learning Object-Agnostic Visual Relationship Features [article]

Xu Yang, Hanwang Zhang, Jianfei Cai
2018 arXiv   pre-print
pairwise patterns, leading to poor generalization to rare or unseen object combinations.  ...  Therefore, we are interested in learning object-agnostic visual features for more generalizable relationship models.  ...  Second, we can exclude the influence of object detection performance, as the improvement of object detection can improve the relationship detection scores [60] .  ... 
arXiv:1808.00171v1 fatcat:qtx5ut5hdjerpgzrb3sxaxseta

Window Detection In Facade Imagery: A Deep Learning Approach Using Mask R-CNN [article]

Nils Nordmark, Mola Ayenew
2021 arXiv   pre-print
We utilize transfer learning to train our proposed method on COCO weights with our own collected dataset of street view images of facades to produce instance segmentations of our new window class.  ...  This article investigates the usage of the mask R-CNN framework to be used for window detection of facade imagery input.  ...  In other words, we are using transfer learning, which means we don't need to train a new model from scratch and that instead utilize a lot of the already learned features from the COCO dataset, which contains  ... 
arXiv:2107.10006v1 fatcat:6m7bgy6rynfoxikz3umeoxn4ce

Learning Coated Adversarial Camouflages for Object Detectors [article]

Yexin Duan, Jialin Chen, Xingyu Zhou, Junhua Zou, Zhengyun He, Jin Zhang, Wu Zhang, Zhisong Pan
2022 arXiv   pre-print
However, the 2D patch attached to a 3D object tends to suffer from an inevitable reduction in attack performance as the viewpoint changes.  ...  Specifically, we make the camouflage perform 3D spatial transformations according to the pose changes of the object.  ...  Overview We aim to generate camouflages which can fool the object detectors to misidentify the object as the target class or hide the object from being detected under different viewing and lighting conditions  ... 
arXiv:2109.00124v3 fatcat:jlsck6nhfvdrrm5gu2rjimc3fu

Learning to Fuse Things and Stuff [article]

Jie Li, Allan Raventos, Arjun Bhargava, Takaaki Tagawa, Adrien Gaidon
2019 arXiv   pre-print
Our model, TASCNet, uses feature maps from a shared backbone network to predict in a single feed-forward pass both things and stuff segmentations.  ...  We propose an end-to-end learning approach for panoptic segmentation, a novel task unifying instance (things) and semantic (stuff) segmentation.  ...  In order to better leverage task affinities and reduce the need for supervision, Zamir et al. [42] build a "taskonomy" by learning general task transfer functions.  ... 
arXiv:1812.01192v2 fatcat:fyvj52qoorg3vntabha7qkoif4

Synthetic Data for Deep Learning [article]

Sergey I. Nikolenko
2019 arXiv   pre-print
Synthetic data is an increasingly popular tool for training deep learning models, especially in computer vision but also in other areas.  ...  Third, we turn to privacy-related applications of synthetic data and review the work on generating synthetic datasets with differential privacy guarantees.  ...  We organize this section by classifying datasets and models with respect to use cases, from generic object detection and segmentation problems to specific domains such as face recognition.  ... 
arXiv:1909.11512v1 fatcat:qquxnw4dfvgmfeztbpdqhr44gy

Cross-View Policy Learning for Street Navigation [article]

Ang Li, Huiyi Hu, Piotr Mirowski, Mehrdad Farajtabar
2019 arXiv   pre-print
Experimental results suggest that the proposed cross-view policy learning enables better generalization of the agent and allows for more effective transfer to unseen environments.  ...  The ability to navigate from visual observations in unfamiliar environments is a core component of intelligent agents and an ongoing challenge for Deep Reinforcement Learning (RL).  ...  The results also suggest that the proposed cross-view learning approach is able to significantly improve the generalization of the representation and the transferability of the street-view agent.  ... 
arXiv:1906.05930v2 fatcat:mwf5brjotzedrk2ll4ibypv5cq

Shuffle-Then-Assemble: Learning Object-Agnostic Visual Relationship Features [chapter]

Xu Yang, Hanwang Zhang, Jianfei Cai
2018 Lecture Notes in Computer Science  
pairwise patterns, leading to poor generalization to rare or unseen object combinations.  ...  Therefore, we are interested in learning object-agnostic visual features for more generalizable relationship models.  ...  Second, we can exclude the influence of object detection performance, as the improvement of object detection can improve the relationship detection scores [60] .  ... 
doi:10.1007/978-3-030-01258-8_3 fatcat:uwxaum5yfjhszjl4cpjjv4ccbe

Learning from Maps: Visual Common Sense for Autonomous Driving [article]

Ari Seff, Jianxiong Xiao
2016 arXiv   pre-print
Experimental evaluation demonstrates that our model learns to correctly infer the road attributes using only panoramas captured by car-mounted cameras as input.  ...  Our goal in this work is to develop a model for road layout inference given imagery from on-board cameras, without any reliance on high-definition maps.  ...  While we developed our model using street view images from one geographic region, it will be interesting to see how well the learned networks can transfer across distant regions (e.g., train on San Francisco  ... 
arXiv:1611.08583v2 fatcat:wes3j26ijbb7vaoupqztq2qtxa
« Previous Showing results 1 — 15 out of 40,157 results