Merge pull request #78 from bkhanal-11/master

Added conclusion in UNet article

Author: vaghawan

_posts/Applied Machine Learning/2022-10-11-semantic-segmentation-unet.md renamed to _posts/Applied Machine Learning/2022-10-12-semantic-segmentation-unet.md

@@ -1,7 +1,7 @@
 ---
 title: "Semantic Segmentation using U-Net"
 excerpt_separator: "<!--more-->"
-last_modified_at: 2022-10-11T14:36:02-05:00
+last_modified_at: 2022-10-12T14:36:02-05:00
 categories:
   - Applied Machine Learning
 tags:
@@ -16,10 +16,10 @@ author: bishwash
 
 ## U-Net
 
-U-Net is a u-shaped encoder-decoder network architecture, which consists of four encoder blocks 
-and four decoder blocks that are connected via bridge. It is one of the most popularly used approaches 
+U-Net is a U-shaped encoder-decoder network architecture, which consists of four encoder blocks 
+and four decoder blocks that are connected via bridges with skip connections. It is one of the most popularly used approaches 
 in any semantic segmentation task. It was originally introduced by Olaf Ronneberger through the publication
-"U-Net: Convolutional Networks for Biomedical Image Segmentation". It is a fully convolutional neural network 
+"U-Net: Convolutional Networks for Biomedical Image Segmentation". It is built upon fully convolutional neural network 
 that is designed to learn from fewer training samples.
 
 <p align="center">
@@ -28,20 +28,20 @@ that is designed to learn from fewer training samples.
 </p>
 
 It has three main componets namely encoder network, decoder network and skip connections. The encoder network 
-(contracting path) halfs the spatial dimensions and doubles the number of feature channels at each encoder block 
-while the decoder network doubles the spatial dimensions and halfs the number of of feature channels. The skip 
+(contracting path) half the spatial dimensions and double the number of feature channels at each encoder block 
+while the decoder network double the spatial dimensions and half the number of of feature channels. The skip 
 connections connect output of encoder block with corresponding input of decoder block.
 
 
 ### Encoder Network
 
 Encoder Network acts as the feature extractor and learn an abstract representation of the input image through 
-a sequence of the encoder blocks. Each encoder block consists of 3x3 convolutions where each convolution is followed by 
+a sequence of the encoder blocks. Each encoder block consists of $3 \times 3$ convolutions where each convolution is followed by 
 a ReLU (Rectified Linear Unit) activation function. The ReLU function introduces non-linearity into the network, which 
 helps in the better generalization of the training data. The output of ReLU acts as a skip connection for the corresponding 
 decoder block.
 
-Next follows a 2x2 max-pooling, where the spatial dimensions of the feature maps are reduced by half. This reduces the 
+Next follows a $2 \times 2$ max-pooling, where the spatial dimensions of the feature maps are reduced by half. This reduces the 
 computational cost by decreasing the number of trainable parameters.
 
 ### Skip Connection
@@ -53,10 +53,10 @@ The bridge connects the encoder and decoder network and completes the flow of in
 
 ### Decoder Network
 
-It is used to take the abstract representation and generate a semantic segmentation mask. The decoder block starts with 2x2 transpose 
+It is used to take the abstract representation and generate a semantic segmentation mask. The decoder block starts with $2 \times 2$ transpose 
 convolution. Next, it is concatenated with the corresponding skip connection feature map from the encoder block. These skip 
 connections provide features from earlier layers that are sometimes lost due to the depth of the network. The output of the last 
-decoder passes through 1x1 convolution with sigmoid activation. The sigmoid activation function gives the segmenation mask representing the 
+decoder passes through $1 \times 1$ convolution with sigmoid activation. The sigmoid activation function gives the segmenation mask representing the 
 pixel-wise classification.
 
 It is prefered to use batch normalization in between the convolution layer and the ReLU activation function. It reduces internal 
@@ -69,22 +69,23 @@ This in turn helps the network to better generalize and prevent it from overfitt
 
 This dataset provides data images and labeled semantic segmentations captured via CARLA self-driving car simulator. The data was 
 generated as part of the Lyft Udacity Challenge . This dataset can be used to train ML algorithms to identify semantic segmentation 
-of cars, roads etc in an image. The data has 5 sets of 1000 images and corresponding labels. There are 23 different labels ranging from 
+of cars, roads etc in an image. The data has $5$ sets of $1000$ images and corresponding labels. There are $23$ different labels ranging from 
 road, roadlines, sidewalk to building, pedestrians, fences. 
 
 <p align="center">
     <img src="{{ site.url }}{{ site.baseurl }}/assets/images/unet/example.png" width="600"/>
 </p>
 
-For our training, we select the first 13 labels. Then all the images were resized to 256x256. The train-validation-test split was 
-0.6, 0.2 and 0.2. Learning rate was chosen to be 0.001 for Adam optimizer. The performance of the network was optimized with the help 
-of Dice Loss which is defined as 
+For our training, we chose the first $13$ labels. All the images and labels were resized to $256 \times 256$ to avoid any tensor mismatch while training. 
+The train-validation-test split was $0.6$, $0.2$ and $0.2$. Learning rate was chosen to be $0.001$ for Adam optimizer. The performance of the 
+network was optimized with the help of Dice Loss which is defined as 
 
 <p align="center">
     <img src="{{ site.url }}{{ site.baseurl }}/assets/images/unet/diceloss.png" width="350"/>
 </p>
 
-The performance of network for for first 25 epoch out of 100 is as follow.
+where $p_{true}, p_{pred}$ are ground truth and predicted labels. $\epsilon$ is small number $\leq 1$. The network was trained over 100 epochs. 
+Other network parameters were {batch_size=8, device=gpu}.The performance of network for the first 25 epoch is given as
 
 <p align="center">
     <img src="{{ site.url }}{{ site.baseurl }}/assets/images/unet/loss.png" width="450"/>
@@ -96,6 +97,13 @@ Some predicted results for buildings with their ground truth is as follow.
     <img src="{{ site.url }}{{ site.baseurl }}/assets/images/unet/results.png" width="600"/>
 </p>
 
+## Conclusion
+
+In all, U-Net works exceptionally for basic semantic segmentation tasks even when we have fewer training samples. 
+Improved version to original U-Net like [U-Net++](https://arxiv.org/abs/1807.10165), 
+[Dense-UNet](https://www.sciencedirect.com/science/article/abs/pii/S0030401821002200) have been introduced for specific tasks 
+in medical and aerial imaging. 
+
 #### References
 
 [1] [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://arxiv.org/pdf/1505.04597.pdf)