Factors of Transferability for a Generic ConvNet Representation

Azizpour, Hossein; Razavian, Ali Sharif; Sullivan, Josephine; Maki, Atsuto; Carlsson, Stefan

doi:10.1109/tpami.2015.2500224

Cited by 280 publications

(274 citation statements)

References 46 publications

(48 reference statements)

Supporting

Mentioning

252

Contrasting

Unclassified

Order By: Relevance

“…Experiments confirm that the top layers may exhibit lower generalization ability than the layer before it. For example, for AlexNet pre-trained on ImageNet, it is shown that FC6, FC7, and FC8 are in descending order regarding retrieval accuracy [130]. It is also shown in [10], [134] that the pool5 feature of AlexNet and VGGNet is even superior to FC6 when proper encoding techniques are employed.…”

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 96%

“…A hybrid dataset combining the Places-205 and the ImageNet datasets has also been used for pre-training [129]. The resulting HybridNet is evaluated in [125], [126], [130], [131] for instance retrieval.…”

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

“…Comprehensive evaluation of various CNNs on instance retrieval has been conducted in several recent works [130], [131], [134]. The transfer effect is mostly concerned.…”

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

“…The transfer effect is mostly concerned. It is considered in [130] that instance retrieval, as a target task, lies farthest from the source, i.e., ImageNet. Studies reveal some critical insights in the transfer process.…”

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

“…Second, the source training set is relevant to retrieval accuracy on different datasets. For example, Azizpour et al [130] report that HybridNet yields the best performance on Holidays after PCA. They also observe that AlexNet pre-trained on ImageNet is superior to PlacesNet and HybridNet on the Ukbench dataset [11] which contains common objects instead of architectures or scenes.…”

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

See 4 more Smart Citations

SIFT Meets CNN: A Decade Survey of Instance Retrieval

Zheng

Yang

Tian

2018

IEEE Trans. Pattern Anal. Mach. Intell.

655

319

View full text Add to dashboard Cite

Abstract-In the early days, content-based image retrieval (CBIR) was studied with global features. Since 2003, image retrieval based on local descriptors (de facto SIFT) has been extensively studied for over a decade due to the advantage of SIFT in dealing with image transformations. Recently, image representations based on the convolutional neural network (CNN) have attracted increasing interest in the community and demonstrated impressive performance. Given this time of rapid evolution, this article provides a comprehensive survey of instance retrieval over the last decade. Two broad categories, SIFT-based and CNN-based methods, are presented. For the former, according to the codebook size, we organize the literature into using large/medium-sized/small codebooks. For the latter, we discuss three lines of methods, i.e., using pre-trained or fine-tuned CNN models, and hybrid methods. The first two perform a single-pass of an image to the network, while the last category employs a patch-based feature extraction scheme. This survey presents milestones in modern instance retrieval, reviews a broad selection of previous works in different categories, and provides insights on the connection between SIFT and CNN-based methods. After analyzing and comparing retrieval performance of different categories on several datasets, we discuss promising directions towards generic and specialized instance retrieval.

show abstract

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 96%

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

“…Comprehensive evaluation of various CNNs on instance retrieval has been conducted in several recent works [130], [131], [134]. The transfer effect is mostly concerned.…”

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

Section: Retrieval Using Pre-trained Cnn Modelsmentioning

confidence: 99%

See 3 more Smart Citations

SIFT Meets CNN: A Decade Survey of Instance Retrieval

Zheng

Yang

Tian

2018

IEEE Trans. Pattern Anal. Mach. Intell.

655

319

View full text Add to dashboard Cite

show abstract

Detection of bodies in maritime rescue operations using unmanned aerial vehicles with multispectral cameras

Gallego

Pertusa

Gil

et al. 2018

Journal of Field Robotics

View full text Add to dashboard Cite

In this study, we use unmanned aerial vehicles equipped with multispectral cameras to search for bodies in maritime rescue operations. A series of flights were performed in open‐water scenarios in the northwest of Spain, using a certified aquatic rescue dummy in dangerous areas and real people when the weather conditions allowed it. The multispectral images were aligned and used to train a convolutional neural network for body detection. An exhaustive evaluation was performed to assess the best combination of spectral channels for this task. Three approaches based on a MobileNet topology were evaluated, using (a) the full image, (b) a sliding window, and (c) a precise localization method. The first method classifies an input image as containing a body or not, the second uses a sliding window to yield a class for each subimage, and the third uses transposed convolutions returning a binary output in which the body pixels are marked. In all cases, the MobileNet architecture was modified by adding custom layers and preprocessing the input to align the multispectral camera channels. Evaluation shows that the proposed methods yield reliable results, obtaining the best classification performance when combining green, red‐edge, and near‐infrared channels. We conclude that the precise localization approach is the most suitable method, obtaining a similar accuracy as the sliding window but achieving a spatial localization close to 1 m. The presented system is about to be implemented for real maritime rescue operations carried out by Babcock Mission Critical Services Spain.

show abstract

A hybrid deep learning architecture for classification of microscopic damage on National Ignition Facility laser optics

Amorin

Kegelmeyer

2019

Statistical Analysis

View full text Add to dashboard Cite

Accurately classifying microscopic damage helps automate the repair and recycling of National Ignition Facility optics and informs the study of damage initiation and growth. This complex 12-class problem previously required human experts to distinguish and label the various damage morphologies. Finding image analysis methods to extract and calculate distinguishing features would be time consuming and challenging, so we sought to automate this task by using convolutional neural networks (CNNs) pretrained on the ImageNet database to take advantage of its automated feature discovery and extraction. We compared three model architectures on this dataset and found the one with highest overall accuracy, 99.17%, was a novel hybrid architecture, one in which we removed the final decision-making layer of the deep learner and replaced it with an ensemble of decision trees (EDT). This combines the power of feature extraction by CNNs with the decision-making strength of EDT.The accuracy of the hybrid architecture over the deep learning alone is shown to be significantly improved. Furthermore, we applied this novel hybrid architecture to an entirely different dataset, one containing images of repaired damage sites, and improved on the previously published findings, also with a demonstrably significant increase in accuracy over using the deep learner alone. K E Y W O R D Sautomation, deep learning, laser optic damage, machine learning

show abstract

Factors of Transferability for a Generic ConvNet Representation

Cited by 280 publications

References 46 publications

SIFT Meets CNN: A Decade Survey of Instance Retrieval

SIFT Meets CNN: A Decade Survey of Instance Retrieval

Detection of bodies in maritime rescue operations using unmanned aerial vehicles with multispectral cameras

A hybrid deep learning architecture for classification of microscopic damage on National Ignition Facility laser optics

Contact Info

Product

Resources

About