Fdflownet: Fast Optical Flow Estimation Using A Deep Lightweight Network

Kong, Lingtong; Yang, Jie

doi:10.1109/icip40778.2020.9191101

Cited by 12 publications

(11 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the detection network can only locate the target in a small area, but not achieve the pixel-level segmentation effect, and it is difficult to present the contour shape completely, affecting the precision of subsequent harvest. Some studies considered that the information distillation mechanism is not efficient and proposed improved recognition methods [35][36][37]. Although the lightweight performance was improved, the overall number of parameters is still not small.…”

Section: Introductionmentioning

confidence: 99%

A Method of Fast Segmentation for Banana Stalk Exploited Lightweight Multi-Feature Fusion Deep Neural Network

Chen¹,

Zhang²,

Zhu³

et al. 2021

Machines

View full text Add to dashboard Cite

In an orchard environment with a complex background and changing light conditions, the banana stalk, fruit, branches, and leaves are very similar in color. The fast and accurate detection and segmentation of a banana stalk are crucial to realize the automatic picking using a banana picking robot. In this paper, a banana stalk segmentation method based on a lightweight multi-feature fusion deep neural network (MFN) is proposed. The proposed network is mainly composed of encoding and decoding networks, in which the sandglass bottleneck design is adopted to alleviate the information a loss in high dimension. In the decoding network, a different sized dilated convolution kernel is used for convolution operation to make the extracted banana stalk features denser. The proposed network is verified by experiments. In the experiments, the detection precision, segmentation accuracy, number of parameters, operation efficiency, and average execution time are used as evaluation metrics, and the proposed network is compared with Resnet_Segnet, Mobilenet_Segnet, and a few other networks. The experimental results show that compared to other networks, the number of network parameters of the proposed network is significantly reduced, the running frame rate is improved, and the average execution time is shortened.

show abstract

Section: Introductionmentioning

confidence: 99%

A Method of Fast Segmentation for Banana Stalk Exploited Lightweight Multi-Feature Fusion Deep Neural Network

Chen¹,

Zhang²,

Zhu³

et al. 2021

Machines

View full text Add to dashboard Cite

show abstract

“…Given input frames of I 1 , I 2 , our OAS-Net estimates optical flow in a coarse-to-fine manner like many approaches [6,8,12,10], where a feature extraction network is adopted to build pyramid features…”

Section: Overall Network Architecturementioning

confidence: 99%

“…Following this, IRR-PWC [9] proposes shared flow estimators among different scales for iterative residual estimation, which can reduce learnable parameters and speed up convergence. FDFlowNet [10] improves original pyramid structure with a compact U-shape network and proposes efficient partial fully connected flow estimator for fast and accurate inference. Recently, Devon [11] builds deformable cost volume with multiple dilated rates to capture small fast moving objects and alleviate warping artifacts.…”

Section: Introductionmentioning

confidence: 99%

“…Though benchmark results are constantly being promoted, there are still challenges blocking further progresses. Since the high efficiency on reducing large displacement, warping operation appears in almost all advanced flow architectures [5,7,6,8,9,12,10]. However, as shown in Fig 1c, the warped target image suffers from ghosting problem, e.g., the yellow cross region, which leads to ambiguity of original scene structure and will further damage performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

OAS-Net: Occlusion Aware Sampling Network for Accurate Optical Flow

Kong

Yang

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

Optical flow estimation is an essential step for many realworld computer vision tasks. Existing deep networks have achieved satisfactory results by mostly employing a pyramidal coarse-to-fine paradigm, where a key process is to adopt warped target feature based on previous flow prediction to correlate with source feature for building 3D matching cost volume. However, the warping operation can lead to troublesome ghosting problem that results in ambiguity. Moreover, occluded areas are treated equally with non occluded regions in most existing works, which may cause performance degradation. To deal with these challenges, we propose a lightweight yet efficient optical flow network, named OAS-Net (occlusion aware sampling network) for accurate optical flow. First, a new sampling based correlation layer is employed without noisy warping operation. Second, a novel occlusion aware module is presented to make raw cost volume conscious of occluded regions. Third, a shared flow and occlusion awareness decoder is adopted for structure compactness. Experiments on Sintel and KITTI datasets demonstrate the effectiveness of proposed approaches.

show abstract

“…The third category explores the motion representation of compressed videos, which takes motion vector and residual error in compression techniques as new modalities of additional stream [19,20,21]. As an important modality in the second category introduced above, optical flow has attracted a lot of attention from researchers, including traditional optimization methods [9], supervised methods trained with pixel-wise labels [9,10,11,12,22,23] and unsupervised approaches trained by brightness consistency [17,13].…”

Section: Introductionmentioning

confidence: 99%

Unsupervised Motion Representation Enhanced Network for Action Recognition

Yang

Kong

Yang

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Learning reliable motion representation between consecutive frames, such as optical flow, has proven to have great promotion to video understanding. However, the TV-L1 method, an effective optical flow solver, is time-consuming and expensive in storage for caching the extracted optical flow. To fill the gap, we propose UF-TSN, a novel end-to-end action recognition approach enhanced with an embedded lightweight unsupervised optical flow estimator. UF-TSN estimates motion cues from adjacent frames in a coarse-to-fine manner and focuses on small displacement for each level by extracting pyramid of feature and warping one to the other according to the estimated flow of the last level. Due to the lack of labeled motion for action datasets, we constrain the flow prediction with multi-scale photometric consistency and edge-aware smoothness. Compared with state-of-the-art unsupervised motion representation learning methods, our model achieves better accuracy while maintaining efficiency, which is competitive with some supervised or more complicated approaches.

show abstract

Fdflownet: Fast Optical Flow Estimation Using A Deep Lightweight Network

Cited by 12 publications

References 13 publications

A Method of Fast Segmentation for Banana Stalk Exploited Lightweight Multi-Feature Fusion Deep Neural Network

A Method of Fast Segmentation for Banana Stalk Exploited Lightweight Multi-Feature Fusion Deep Neural Network

OAS-Net: Occlusion Aware Sampling Network for Accurate Optical Flow

Unsupervised Motion Representation Enhanced Network for Action Recognition

Contact Info

Product

Resources

About