In general, CNN for the semantic segmentation consists of two essential parts for the pixel-wise classification to assign the pixel labels. The first part is encoder composed of convolutional, which is responsible for the automatic features extraction and sub-sampling layers. The purpose of the sub-sampling layers is to achieve spatial in-variance by reducing the resolution of the feature maps, which also improves the robustness of the classifier due to the elimination of the redundant features. Sub-sampling also increases the field of view of the feature maps to extract more abstract class salient features and minimizes computation time. Sometimes, dropout layer or batch normalization or even both are added after each convolution to avoid the overfitting and to make the robust pixel-wise classifier. On the other hand, the second part is decoder to semantically project the discriminating features of lower resolution learned by the encoder onto the pixel space of higher resolution to get a dense pixel-wise classification. In semantic segmentation, the encoder part is quite similar to all the CNN models but they mainly differ in decoder mechanism. Semantic segmentation not only requires discrimination at pixel level but also a decoder mechanism to project the discriminating features learned at different stages of the encoder onto the pixel space. However, the significantly reduced features map due to sub-sampling suffers spatial resolution loss, which introduces coarseness, less edge information, checkerboard noise, and over-segmentation in the semantically segmented image. When the kernel size of the deconvolution is not divisible by the up-scaling factor, the number of low-resolution features that contribute to a single high-resolution feature is not constant across the high-resolution feature maps, which is called deconvolution overlap and is one of the causes of checkerboard artifact in the segmented mask.
-
Title: How to do Semantic Segmentation using Deep learning
Link : https://medium.com/nanonets/how-to-do-image-segmentation-using-deep-learning-c673cc5862ef -
Title: Semantic Segmentation using CNN’s
Link : https://medium.com/@aiclubiiitb/semantic-segmentation-using-cnns-357fbcfa1bc -
Title: Semantic Segmentation with Deep Learning
Link : https://towardsdatascience.com/semantic-segmentation-with-deep-learning-a-guide-and-code-e52fc8958823 -
Title: Semantic Segmentation: Introduction to the Deep Learning Technique Behind Google Pixel’s Camera!
Link : https://www.analyticsvidhya.com/blog/2019/02/tutorial-semantic-segmentation-google-deeplab/ -
Title: Computer Vision Tutorial: A Step-by-Step Introduction to Image Segmentation Techniques
Link : https://www.analyticsvidhya.com/blog/2019/04/introduction-image-segmentation-techniques-python/ -
Title: Semantic Segmentation of Aerial images Using Deep Learning
Link : https://towardsdatascience.com/semantic-segmentation-of-aerial-images-using-deep-learning-90fdf4ad780 -
Title: How to use DeepLab in TensorFlow for object segmentation using Deep Learning
Link : https://medium.freecodecamp.org/how-to-use-deeplab-in-tensorflow-for-object-segmentation-using-deep-learning-a5777290ab6b
-
Title: U-Net: Convolutional Networks for Biomedical Image Segmentation
Link : https://arxiv.org/abs/1505.04597 -
Title: Understanding Semantic Segmentation with UNET
Link : https://towardsdatascience.com/understanding-semantic-segmentation-with-unet-6be4f42d4b47 -
Title: Medical Image Segmentation [Part 1] — UNet: Convolutional Networks with Interactive Code
Link : https://towardsdatascience.com/medical-image-segmentation-part-1-unet-convolutional-networks-with-interactive-code-70f0f17f46c6 -
Title: U-Net
Link : http://www.deeplearning.net/tutorial/unet.html -
Title: Learn How to Train U-Net On Your Dataset
Link : https://medium.com/coinmonks/learn-how-to-train-u-net-on-your-dataset-8e3f89fbd623 -
Title: Practical image segmentation with Unet
Link : https://tuatini.me/practical-image-segmentation-with-unet/ -
Title: Get acquainted with U-NET architecture + some keras shortcuts
Link : https://spark-in.me/post/unet-adventures-part-one-getting-acquainted-with-unet
-
Title: Fully Convolutional Networks (FCN) for 2D segmentation
Link : http://www.deeplearning.net/tutorial/fcn_2D_segm.html -
Title: Review: FCN — Fully Convolutional Network (Semantic Segmentation)
Link : https://towardsdatascience.com/review-fcn-semantic-segmentation-eb8c9b50d2d1 -
Title: Fully Convolutional Networks (FCNs) for Image Segmentation
Link : http://warmspringwinds.github.io/tensorflow/tf-slim/2017/01/23/fully-convolutional-networks-(fcns)-for-image-segmentation/ -
Title: Semantic Segmentation using Fully Convolutional Networks over the years
Link : https://meetshah1995.github.io/semantic-segmentation/deep-learning/pytorch/visdom/2017/06/01/semantic-segmentation-over-the-years.html -
Title: Fully Convolutional Networks for Semantic Segmentation
Link : https://paperswithcode.com/paper/fully-convolutional-networks-for-semantic -
Title: Fully Convolutional Networks (FCN)
Link : https://d2l.ai/chapter_computer-vision/fcn.html -
Title: How is Fully Convolutional Network (FCN) different from the original Convolutional Neural Network (CNN)?
Link : https://www.quora.com/How-is-Fully-Convolutional-Network-FCN-different-from-the-original-Convolutional-Neural-Network-CNN
- Title: Awesome Semantic Segmentation
Link : https://github.com/mrgloom/awesome-semantic-segmentation
-
Title: The International Skin Imaging Collaboration (ISIC) is an international effort to improve melanoma diagnosis, sponsored by the International Society for Digital Imaging of the Skin (ISDIS). The ISIC Archive contains the largest publicly available collection of quality controlled dermoscopic images of skin lesions.
Link : https://www.isic-archive.com/#!/topWithHeader/wideContentTop/main -
Title: All challenges that have been organized by MICCAI within the area of medical image analysis can be found on the following link.
Link : https://grand-challenge.org/challenges/In the picture, it is seen that after selecting task type as Segmentation, the Medical image segmentation challenges can be found.
-
More other challenges:
Link: http://www.isles-challenge.org/
Link: http://atriaseg2018.cardiacatlas.org/
Link: http://medicaldecathlon.com/
Link: http://braintumorsegmentation.org/
Link: http://sceneparsing.csail.mit.edu/
Link: http://cocodataset.org/#home