Mask Encoding for Single Shot Instance Segmentation

Zhang, Rufeng; Tian, Zhi; Shen, Chunhua; You, Mingyu; Yan, Youliang

doi:10.1109/cvpr42600.2020.01024

Cited by 100 publications

(45 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second measurement in the boundary detection called as Measure II gauges the proportion of mask pixels among predesignated areas. We compare Measure II in both models with algae and cell membrane images, where the Mask R-CNN produces overlapping inferred masks of two [33]) models via Measure I and present the mean and standard errors given the prespecified number of circles and ellipses (i.e., 4 and 6) in Tables 1 and 2. The results indicate that the predicted mask of the weighted Mask R-CNN model is superior across simulation scenarios when we estimate the ground truth mask compared to the Mask R-CNN and MEInst.…”

Section: Journal Of Sensorsmentioning

confidence: 99%

Weighted Mask R‐CNN for Improving Adjacent Boundary Segmentation

Suh

Park

et al. 2021

Journal of Sensors

View full text Add to dashboard Cite

In the recent era of AI, instance segmentation has significantly advanced boundary and object detection especially in diverse fields (e.g., biological and environmental research). Despite its progress, edge detection amid adjacent objects (e.g., organism cells) still remains intractable. This is because homogeneous and heterogeneous objects are prone to being mingled in a single image. To cope with this challenge, we propose the weighted Mask R-CNN designed to effectively separate overlapped objects in virtue of extra weights to adjacent boundaries. For numerical study, a range of experiments are performed with applications to simulated data and real data (e.g., Microcystis, one of the most common algae genera and cell membrane images). It is noticeable that the weighted Mask R-CNN outperforms the standard Mask R-CNN, given that the analytic experiments show on average 92.5% of precision and 96.4% of recall in algae data and 94.5% of precision and 98.6% of recall in cell membrane data. Consequently, we found that a majority of sample boundaries in real and simulated data are precisely segmented in the midst of object mixtures.

show abstract

Section: Journal Of Sensorsmentioning

confidence: 99%

Weighted Mask R‐CNN for Improving Adjacent Boundary Segmentation

Suh

Park

et al. 2021

Journal of Sensors

View full text Add to dashboard Cite

show abstract

“…The results demonstrate that our IFR is robust and also flexible for one-stage approaches that directly outputs the full instance masks from whole features without RoIAlign. For instance, our proposed IFR improves the average AP by +1.0% mAP on MEInst [48]. Similarly, more improvements on large-scale objects are achieved due to the enlarged receptive field.…”

Section: Methodsmentioning

confidence: 77%

Implicit Feature Refinement for Instance Segmentation

Wang²,

Dong³

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

We propose a novel implicit feature refinement module for highquality instance segmentation. Existing image/video instance segmentation methods rely on explicitly stacked convolutions to refine instance features before the final prediction. In this paper, we first give an empirical comparison of different refinement strategies, which reveals that the widely-used four consecutive convolutions are not necessary. As an alternative, weight-sharing convolution blocks provides competitive performance. When such block is iterated for infinite times, the block output will eventually converge to an equilibrium state. Based on this observation, the implicit feature refinement (IFR) is developed by constructing an implicit function. The equilibrium state of instance features can be obtained by fixed-point iteration via a simulated infinite-depth network. Our IFR enjoys several advantages: 1) simulates an infinite-depth refinement network while only requiring parameters of single residual block; 2) produces high-level equilibrium instance features of global receptive field; 3) serves as a plug-and-play general module easily extended to most object recognition frameworks. Experiments on the COCO and YouTube-VIS benchmarks show that our IFR achieves improved performance on state-of-the-art image/video instance segmentation frameworks, while reducing the parameter burden (e.g. 1% AP improvement on Mask R-CNN with only 30.0% parameters in mask head). Code is made available at https://github.com/lufanma/IFR.git.

show abstract

“…Segmentation in general, and instance segmentation in particular, has benefited strongly from the adoption of CNNs to increase performance, which has led to the proposal of multiple models in very recent years [45][46][47][48][49][50][51]. This study aims at mice pupil segmentation in images acquired in a low-light environment.…”

Section: Deep Learning-based Methods For Pupil Segmentationmentioning

confidence: 99%

Automated Mouse Pupil Size Measurement System to Assess Locus Coeruleus Activity with a Deep Learning-Based Approach

Lara-Doña

Torres-Sánchez

Torres

et al. 2021

Sensors

View full text Add to dashboard Cite

Strong evidence from studies on primates and rodents shows that changes in pupil diameter may reflect neural activity in the locus coeruleus (LC). Pupillometry is the only available non-invasive technique that could be used as a reliable and easily accessible real-time biomarker of changes in the in vivo activity of the LC. However, the application of pupillometry to preclinical research in rodents is not yet fully standardized. A lack of consensus on the technical specifications of some of the components used for image recording or positioning of the animal and cameras have been recorded in recent scientific literature. In this study, a novel pupillometry system to indirectly assess, in real-time, the function of the LC in anesthetized rodents is presented. The system comprises a deep learning SOLOv2 instance-based fast segmentation framework and a platform designed to place the experimental subject, the video cameras for data acquisition, and the light source. The performance of the proposed setup was assessed and compared to other baseline methods using a validation and an external test set. In the latter, the calculated intersection over the union was 0.93 and the mean absolute percentage error was 1.89% for the selected method. The Bland–Altman analysis depicted an excellent agreement. The results confirmed a high accuracy that makes the system suitable for real-time pupil size tracking, regardless of the pupil’s size, light intensity, or any features typical of the recording process in sedated mice. The framework could be used in any neurophysiological study with sedated or fixed-head animals.

show abstract

Mask Encoding for Single Shot Instance Segmentation

Cited by 100 publications

References 26 publications

Weighted Mask R‐CNN for Improving Adjacent Boundary Segmentation

Weighted Mask R‐CNN for Improving Adjacent Boundary Segmentation

Implicit Feature Refinement for Instance Segmentation

Automated Mouse Pupil Size Measurement System to Assess Locus Coeruleus Activity with a Deep Learning-Based Approach

Contact Info

Product

Resources

About