AdapNet: Adaptability Decomposing Encoder–Decoder Network for Weakly Supervised Action Recognition and Localization

Zhang, Xiaoyu; Li, Changsheng; Shi, Haichao; Zhu, Xiaobin; Li, Peng

doi:10.1109/tnnls.2019.2962815

Cited by 39 publications

(17 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the rapid development of deep learning, many deep neural networks have been proposed, such as AlexNet [39], ZF [40], VGG [36], GoogleNet [41], and ResNet [42], etc. Different network frameworks and learning strategies had been designed in some researches, which effectively improve the generalization ability of the network models [43]- [45]. By designing different numbers of weight layers, networks with different depths can be built by these networks.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Approach for Apple Edge Detection to Remotely Monitor Apple Growth in Orchards

Wang

Song

et al. 2020

IEEE Access

View full text Add to dashboard Cite

The automatic intelligent acquisition of apple growth information in the long-term provides a promising benefit for growers to plan the application of nutrients and pesticides during apple maturation. The overall goal of this study was to develop an apple growth monitoring system in an orchard based on a deep learning edge detection network for apple size remote estimation throughout the entire growth period. A remote apple growth monitoring hardware system was built with a spherical video camera and two personal computers to regularly acquire apple images. For software, an edge detection network that fused convolutional features (FCF) was proposed to segment the apple images. To filter out irrelevant apples in the images, points on apples to be monitored were manually selected from the images as seed points, and the region growing method was conducted on the extracted edge maps. Then, the horizontal diameters of the apples were calculated. The experimental results showed that the F1 score of the FCF method was 53.1% on the apple test set, and the average run time was 0.075 s per image, which was better than the other five methods in comparison. The growth of the apples was monitored by our system from the date after apple thinning to apple ripening. The mean average absolute error of the apples' horizontal diameters detected by our system was 0.90 mm, and it decreased by 67.9% when compared with the circle fitting-based method (2.8 mm). These results suggest that our system provides an effective and accurate way to monitor the growth of apples on the trees. The proposed method provides a reference for monitoring the growth of other fruits during the growth period, and it can be used to optimize orchard management.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep Learning Approach for Apple Edge Detection to Remotely Monitor Apple Growth in Orchards

Wang

Song

et al. 2020

IEEE Access

View full text Add to dashboard Cite

show abstract

“…All the previously mentioned works propose fully supervised approaches. Since it is more complicated each day to have labels for such big amount of videos, the community is exploring also weakly supervised alternatives [22]- [29]. In any case, our analysis focuses on the different problem of online action detection.…”

Section: Related Workmentioning

confidence: 99%

Rethinking Online Action Detection in Untrimmed Videos: A Novel Online Evaluation Protocol

et al. 2020

View full text Add to dashboard Cite

The Online Action Detection (OAD) problem needs to be revisited. Unlike traditional offline action detection approaches, where the evaluation metrics are clear and well established, in the OAD setting we find very few works and no consensus on the evaluation protocols to be used. In this work we propose to rethink the OAD scenario, clearly defining the problem itself and the main characteristics that the models which are considered online must comply with. We also introduce a novel metric: the Instantaneous Accuracy (IA). This new metric exhibits an online nature and solves most of the limitations of the previous metrics. We conduct a thorough experimental evaluation on 3 challenging datasets, where the performance of various baseline methods is compared to that of the state-of-the-art. Our results confirm the problems of the previous evaluation protocols, and suggest that an IA-based protocol is more adequate to the online scenario. The baselines models and a development kit with the novel evaluation protocol are publicly available here (https://github.com/gramuah/ia).

show abstract

“…In particular, they aggregated activations of each layer and concatenated them into the final representation, which achieved satisfactory results. The research of [26] and [27] also showed that visual recognition tasks make a considerable difference, which needs to be considered in the process of constructing a depth model. For example, during the construction of the action recognition model of [28], the adaptability of the model to a weak supervised dataset was taken into account.…”

Section: Related Workmentioning

confidence: 99%

Image Representation Method Based on Relative Layer Entropy for Insulator Recognition

Zhao

Fan

et al. 2020

Entropy

View full text Add to dashboard Cite

Deep convolutional neural networks (DCNNs) with alternating convolutional, pooling and decimation layers are widely used in computer vision, yet current works tend to focus on deeper networks with many layers and neurons, resulting in a high computational complexity. However, the recognition task is still challenging for insufficient and uncomprehensive object appearance and training sample types such as infrared insulators. In view of this, more attention is focused on the application of a pretrained network for image feature representation, but the rules on how to select the feature representation layer are scarce. In this paper, we proposed a new concept, the layer entropy and relative layer entropy, which can be referred to as an image representation method based on relative layer entropy (IRM_RLE). It was designed to excavate the most suitable convolution layer for image recognition. First, the image was fed into an ImageNet pretrained DCNN model, and deep convolutional activations were extracted. Then, the appropriate feature layer was selected by calculating the layer entropy and relative layer entropy of each convolution layer. Finally, the number of the feature map was selected according to the importance degree and the feature maps of the convolution layer, which were vectorized and pooled by VLAD (vector of locally aggregated descriptors) coding and quantifying for final image representation. The experimental results show that the proposed approach performs competitively against previous methods across all datasets. Furthermore, for the indoor scenes and actions datasets, the proposed approach outperforms the state-of-the-art methods.Entropy 2020, 22, 419 2 of 17 efficiency for real-time intelligent insulator recognition methods [6,7]. Figure 1 shows insulators on a transmission line.

show abstract

AdapNet: Adaptability Decomposing Encoder–Decoder Network for Weakly Supervised Action Recognition and Localization

Cited by 39 publications

References 58 publications

Deep Learning Approach for Apple Edge Detection to Remotely Monitor Apple Growth in Orchards

Deep Learning Approach for Apple Edge Detection to Remotely Monitor Apple Growth in Orchards

Rethinking Online Action Detection in Untrimmed Videos: A Novel Online Evaluation Protocol

Image Representation Method Based on Relative Layer Entropy for Insulator Recognition

Contact Info

Product

Resources

About