Eff-UNet: A Novel Architecture for Semantic Segmentation in Unstructured Environment

Baheti, Bhakti; Innani, Shubham; Gajre, Suhas; Talbar, Sanjay N.

doi:10.1109/cvprw50498.2020.00187

Cited by 184 publications

(90 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…[67] that infers the PLM counterparts of the input SLIM images without the presence of histological staining. The main difference between the U-Net and the E-U-Net is that E-U-Net uses a more efficient pre-trained encoder, EfficientNet, in the encoding path [55]. EfficientNet is a compact convolutional neural network that retains its very high performance while lowering the number of total network parameters being used in order to increase the convergence rate.…”

Section: Efficient-unet Network and Trainingmentioning

confidence: 99%

“…In order to detect solely the collagen fibers from the SLIM image, we trained a convolutional neural network with SLIM as input and PLM images as the ground-truth. We used a modified U-Net, referred to as the Efficient U-Net (E-U-Net) [55]. The encoding path of E-U-Net is the EfficientNet in which the network was pretrained on image dataset from ImageNet…”

Section: Efficient U-netmentioning

confidence: 99%

“…Branching off from the EfficientNet B0, the least complex, more complex architectures were built ranging from B1 to B7 by adding more layers for higher performance [55]. We empirically selected the best performing EfficientNet architecture, EfficientNet B5, as the pre-encoder of our E-U-Net training session by juxtaposing the minimum loss function value attained by each architecture.…”

Section: Efficient-unet Network and Trainingmentioning

confidence: 99%

See 2 more Smart Citations

Label-free imaging of collagen fibers in tissue slices using phase imaging with computational specificity

Sakakura

Popescu

Kajdacsy-Balla

et al. 2021

Preprint

View full text Add to dashboard Cite

Evaluating the tissue collagen content in addition to the epithelial morphology has been proven to offer complementary information in histopathology, especially in disease stratification and patient survivability prediction. One imaging modality widely used for this purpose is second harmonic generation microscopy (SHGM), which reports on the nonlinear susceptibility associated with the collagen fibers. Another method is polarization light microscopy (PLM) combined with picrosirius-red (PSR) tissue staining. However, SHGM requires expensive equipment and provides limited throughput, while PLM and PSR staining are not part of the routine pathology workflow. Here, we advance phase imaging with computational specificity (PICS) to computationally infer the collagen distribution of unlabeled tissue, with high specificity. PICS utilizes deep learning to translate quantitative phase images (QPI) into corresponding PSR images with high accuracy and speed. Our results indicate that the distributions of collagen fiber orientation, length, and straightness reported by PICS closely match the ones from ground truth.

show abstract

Section: Efficient-unet Network and Trainingmentioning

confidence: 99%

Section: Efficient U-netmentioning

confidence: 99%

Section: Efficient-unet Network and Trainingmentioning

confidence: 99%

See 1 more Smart Citation

Label-free imaging of collagen fibers in tissue slices using phase imaging with computational specificity

Sakakura

Popescu

Kajdacsy-Balla

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…To build a smart system, cameras and vision systems play the role of human eyes for mobile robots and the intelligent reasoning/decision part, playing the role of the human brain, include neural network models and processors. Autonomous mobile robots are being developed with more safe and robust features which are based on object detection and semantic segmentation in the surrounding area [3]. Semantic segmentation is the process of classifying each pixel in an image and assigning a predefined set of classes or labels to each pixel which also is called pixel level classification.…”

Section: Introductionmentioning

confidence: 99%

“…In recent years, new models are presented to address the degradation problem to achieve higher accuracies with large amounts of convolutional layers. Although some of these models can be used for better results [2], [3], [8], [9], achieving a proper trade-off between efficiency and accuracy is a challenging task for big datasets. Reaching a high accuracy by increasing convolutional layers causes a high increase in the required computational resources which are a big limitation for fast and real-time applications [2].…”

Section: Introductionmentioning

confidence: 99%

Ghost-UNet: An Asymmetric Encoder-Decoder Architecture for Semantic Segmentation From Scratch

2021

View full text Add to dashboard Cite

One of the most important key points in the intelligent transportation systems is scene understanding of the known and unknown surrounding environment to achieve a safe driving for smart mobile robots and cars. Semantic segmentation can address most of the perception needs of mobile robots and Intelligent Vehicles (IV). There are several deep learning approaches based on Convolutional Neural Network (CNN) for semantic segmentation. Most of these techniques have been designed on a pretrained network base and loading a specific weight file is necessary for them. In this paper, we propose a deep architecture for semantic segmentation from scratch based on an asymmetry encoder-decoder architecture using Ghost-Net and U-Net which we have called it Ghost-UNet. This model can be used for precise segmentation using a combination of low-level spatial information and high-level feature maps. We focus our work on outdoor datasets to evaluate the proposed model which is tested on the Cityscapes dataset. The proposed model has good pixel accuracy and mean Intersection over Union (mIoU) compared with other valid literature.

show abstract

Classification and understanding of cloud structures via satellite images with EfficientUNet

Ahmed¹,

Sabab²

2021

Preprint

View full text Add to dashboard Cite

Climate change has been a common interest and the forefront of crucial political discussion and decision-making for many years. Shallow clouds play a significant role in understanding the Earth's climate but they are challenging to interpret and represent in a climate model. By classifying these cloud structures there is a better possibility of understanding the physical structures of the clouds which would improve the climate model generation resulting in a better prediction of climate change or forecasting weather update. Clouds organise in many forms which makes it challenging to build traditional rule-based algorithms to separate cloud features. In this paper classification of cloud organization patterns was performed using a new scaled-up version of Convolutional Neural Network (CNN) named as EfficientNet as the encoder and UNet as decoder where they worked as feature extractor and reconstructor of fine grained feature map and was used as a classifier which will help experts to understand how clouds will shape the future climate. By using a segmentation model in a classification task it was shown that with a good encoder alongside UNet it is possible to obtain good performance from this dataset. Dice coefficient has been used for the final evaluation metric which gave the score of 66.26% and 66.02% for public and private (test set) leaderboard on Kaggle competition respectively.

show abstract

Eff-UNet: A Novel Architecture for Semantic Segmentation in Unstructured Environment

Cited by 184 publications

References 20 publications

Label-free imaging of collagen fibers in tissue slices using phase imaging with computational specificity

Label-free imaging of collagen fibers in tissue slices using phase imaging with computational specificity

Ghost-UNet: An Asymmetric Encoder-Decoder Architecture for Semantic Segmentation From Scratch

Classification and understanding of cloud structures via satellite images with EfficientUNet

Contact Info

Product

Resources

About