Weighted Mask R‐CNN for Improving Adjacent Boundary Segmentation

Suh, SungMin; Park, Yongeun; Ko, Kyoungmin; Yang, SeongMin; Ahn, Jaehyeong; Shin, Jae-Ki; Kim, Sunghwan

doi:10.1155/2021/8872947

Cited by 19 publications

(13 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in Figure 3G-J, the addition of PEN does not qualitatively improve the segmentation ability of Mask-RCNN compared to training with MIP inputs. This is consistent with previous reports showing Mask-RCNN often struggles in cases of overlapping instances [28], as proposed regions in Mask-RCNN during inference are reduced using non-maximum suppression to prevent multiple detections of the same instance.…”

Section: Proposed Workflowsupporting

confidence: 92%

The Projection-Enhancement Network (PEN)

Eddy¹,

Naylor²,

Sun³

2023

Preprint

View full text Add to dashboard Cite

Contemporary approaches to instance segmentation in cell science use 2D or 3D convolutional networks depending on the experiment and data structures. However, limitations in microscopy systems or efforts to prevent phototoxicity commonly require recording sub-optimally sampled data regimes that greatly reduces the utility of such 3D data, especially in crowded environments with significant axial overlap between objects. In such regimes, 2D segmentations are both more reliable for cell morphology and easier to annotate. In this work, we propose the Projection Enhancement Network (PEN), a novel convolutional module which processes the sub-sampled 3D data and produces a 2D RGB semantic compression, and is trained in conjunction with an instance segmentation network of choice to produce 2D segmentations. Our approach combines augmentation to increase cell density using a low-density cell image dataset to train PEN, and curated datasets to evaluate PEN. We show that with PEN, the learned semantic representation in CellPose encodes depth and greatly improves segmentation performance in comparison to maximum intensity projection images as input, but does not similarly aid segmentation in region-based networks like Mask-RCNN. Finally, we dissect the segmentation strength against cell density of PEN with CellPose on disseminated cells from side-by-side spheroids. We present PEN as a data-driven solution to form compressed representations of 3D data that improve 2D segmentations from instance segmentation networks.

show abstract

Section: Proposed Workflowsupporting

confidence: 92%

The Projection-Enhancement Network (PEN)

Eddy¹,

Naylor²,

Sun³

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…This stage describes the mask R-CNN architecture used for cow segmentation by applying the same network architecture defined in [12], [25]. Figure 3 shows the 2-stages mask R-CNN framework, the first entails scanning the image and generating a region of interest (ROI) where possible objects are located.…”

Section: Mask R-cnnmentioning

confidence: 99%

“…For example, the potential of deep neural networks in feature learning [9], [10] has supported advancements in computer vision, object detection, and segmentation. Regarding object detection, the methods of faster region-based convolutional neural network (R-CNN) [11] and mask R-CNN [12] have contributed significantly to powerful object detection. Also, in object segmentation, the mask R-CNN method has good image object detection and segmentation tactics [13] but has not been able to fully distinguish between foreground and background images.…”

Section: Introductionmentioning

confidence: 99%

Two-stages of segmentation to improve accuracy of deep learning models based on dairy cow morphology

Siregar¹,

Purwanto²,

Wijaya³

et al. 2023

IJECE

View full text Add to dashboard Cite

<span lang="EN-US">Computer vision deals with image-based problems, such as deep learning, classification, and object detection. This study classifies the quality of dairy parents into three, namely high, medium, and low based on morphology by focusing on Bogor Indonesia farms. The morphological images used are the side and back of dairy cows and the challenge is to determine the optimal accuracy of the model for it to be implemented into an automated system. The 2-step mask region-based convolutional neural network (mask R-CNN) and Canny segmentation algorithm were continuously used to classify the convolutional neural network (CNN) in order to obtain optimal accuracy. When testing the model using training and testing ratios of 90:10 and 80:20, it was evaluated in terms of accuracy, precision, recall, and F1-score. The results showed that the highest model produced an accuracy of 85.44%, 87.12% precision, 83.79% recall, and 84.94% F1-score. Therefore, it was concluded that the test result with 2-stages of segmentation was the best.</span>

show abstract

“…In the decoder, the high-frequency module (HFM) enhances high-frequency components to highlight fine-grained information for accurate prediction. Besides, we take the strategy that jointly learning object boundaries and masks in an end-to-end manner [75,26,9,44]. With the interaction between mask branch and boundary branch features, the network can better perceive the localization and shape information, which also helps identify the target objects.…”

Section: Overviewmentioning

confidence: 99%

Global Spectral Filter Memory Network for Video Object Segmentation

Liu¹,

Rong²,

Wang³

et al. 2022

Preprint

View full text Add to dashboard Cite

This paper studies semi-supervised video object segmentation through boosting intra-frame interaction. Recent memory networkbased methods focus on exploiting inter-frame temporal reference while paying little attention to intra-frame spatial dependency. Specifically, these segmentation model tends to be susceptible to interference from unrelated nontarget objects in a certain frame. To this end, we propose Global Spectral Filter Memory network (GSFM), which improves intraframe interaction through learning long-term spatial dependencies in the spectral domain. The key components of GSFM is 2D (inverse) discrete Fourier transform for spatial information mixing. Besides, we empirically find low frequency feature should be enhanced in encoder (backbone) while high frequency for decoder (segmentation head). We attribute this to semantic information extracting role for encoder and fine-grained details highlighting role for decoder. Thus, Low (High) Frequency Module is proposed to fit this circumstance. Extensive experiments on the popular DAVIS and YouTube-VOS benchmarks demonstrate that GSFM noticeably outperforms the baseline method and achieves state-of-theart performance. Besides, extensive analysis shows that the proposed modules are reasonable and of great generalization ability. Our source code is available at https://github.com/workforai/GSFM.

show abstract

Weighted Mask R‐CNN for Improving Adjacent Boundary Segmentation

Cited by 19 publications

References 35 publications

The Projection-Enhancement Network (PEN)

The Projection-Enhancement Network (PEN)

Two-stages of segmentation to improve accuracy of deep learning models based on dairy cow morphology

Global Spectral Filter Memory Network for Video Object Segmentation

Contact Info

Product

Resources

About