We’re rewarding the question askers & reputations are being recalculated! Read more.

New answers tagged

0

Problem is in the output layer and you are using categorical_crossentropy for a loss function. Quoting Keras documentation: Note: when using the categorical_crossentropy loss, your targets should be in categorical format (e.g. if you have 10 classes, the target for each sample should be a 10-dimensional vector that is all-zeros except for a 1 at the ...


0

Chapter 9 of the book Deep Learning (2016), by Goodfellow et al., describes the convolutional (neural) network (CNN), its main operations (namely, convolution and pooling) and properties (such as parameter sharing). There's also the article From Convolution to Neural Network, which first introduces the mathematical operation convolution and then describes ...


0

I'm not sure if this is what you are looking for but I find Goodfellow's book a pretty good resource: Goodfellow: https://www.deeplearningbook.org/ 'Pattern Recognition and Machine Learning' by Bishop Might also be useful. If you edit your question to post a bit more detail, we can offer better answers, for example, what is about the convolutional layer? ...


0

Chris Olah's work is always inspired, and not too technical as one would expect. He has several papers on CNNs on his website. In particular, check the series titled "Convolutional Neural Networks" with four papers on the topic.


1

Yes you can, a few years ago I made a simple CNN for a single Arabic phoneme classification. You can use spectogram or using MFCC / MFSC as features, as long all data has the same size (use padding or cropping if needed). You may need RNN if you want to combine some phonemes to recognize a single word or longer.


2

This is a really cool problem. You already have a working model here are a few different ways of going forward with the project. Grouping text based on locality. "no segmentation" Text region extraction in a document image based on the Delaunay tessellation Segmentation Multiscale Edge-Based Text Extraction from Complex Images Training a map of the ...


2

I'm working on a similar problem. I'm using a 2D point cloud of an object, for example, X and Y coordinates for height, and with that more simple data set I will train a regression model (currently working on that). In my opinion, this approach with dissecting complex point cloud into cross sections that contain wanted dimension and feeding that to the model ...


0

The min and max input size should be the min and max image size of the input images. The numbers represent pixels in both axis of the image. The anchors represents the size of the anchor box. Anchor box does not have coordinates, only have size. Hope I can help you


1

First of all, you should add the argument workers = n in the fit generator call. n should be bigger than 1 to prefetch data. As your data processing requires the data be taken from a server or port, you should do pre fetching data as that would fetch the next data while GPU is processing. If you call fit_generator with workers > 1 , use_multiprocessing=...


0

The answer is yes this is possible and here are the papers where they do almost exactly the same project you are describing above. Although none of the bellow combine gazebo, single point/single shot, 6D-pose and CNNs. In order to use synthetic data to train a model that works on real data. Pose Estimation by Key Points Registration in Point Cloud (2019) ...


0

Spectral Convolution In a spectral graph convolution, we perform an Eigen decomposition of the Laplacian Matrix of the graph. This Eigen decomposition helps us in understanding the underlying structure of the graph with which we can identify clusters/sub-groups of this graph. This is done in the Fourier space. An analogy is PCA where we understand the spread ...


0

You can use this to view the Keras Resnet Inception V2 network. from keras.applications.inception_resnet_v2 import InceptionResNetV2, preprocess_input from keras.layers import Input model = InceptionResNetV2(weights='imagenet', include_top=True) print(model.summary()) This will Output (im showing only the last few layers): ...


1

Read on Fully Convolutional Networks (FCN). There is a lot of papers on the subject, first was "Fully Convolutional Networks for Semantic Segmentation" by Long. The idea is quite close to what you describe - preserve spatial locality in the layers. In FCN there is no fully connected layer. Instead there is average pooling on top of last low-resolution/high-...


3

I have had similar thoughts about neural networks before. Convolution layers are layers of two dimensional nodes effectively passing the spacial data so why don't we use two dimensional hidden layers to receive information out of them. I'm sure someone has used this type of implementation before. I believe the papers bellow are using this. Part of the ...


1

Other replies are commenting on the skip connections for a U-Net. I believe you want to exclude these skip connections from your auto-encoder. You say you want to use the auto-encoder for unsupervised pretraining, for which you want to pass the data through a bottle neck, so adding skip connections would work against you if you want to use the encoder for a ...


1

Personally, I'd say as long as the object is visible don't do either. If the model has been well built and if lighting changes would help, the convolution operation weights would learn an operation similar to contrast or brightness changes. On the other hand if the object visibility is an issue, then natural lighting changes would be better, due to the lack ...


2

It depends on what is your ultimate goal. If your goal is to simply classify the object in the image, having more complex output won't help. Simpler output representation yields better result. If your goal is to detect the bounding box, output the bounding box. There is no need for a more complex output feature. If you use a segmentation method for bounding ...


1

Yes you can vectorize a CNN. See this github file for details: https://github.com/parasdahal/deepnet/blob/master/deepnet/layers.py After looking through it it basically transposes the input to some dimension and apply matrix multiplication to the weight with some other kind of transfromation. Pls refer to the github repository for details. Hope this can ...


1

If you want to count the number of objects using a neural network, you can use pretrained YOLO with the bottom prediction layer removed, and feed the features to a classification feed forward layer of let's say 1000 class representing 0-999 objects in the image. You can then train it and propagate the gradients through it. For example, in the pytorch code ...


2

For extra input that does not matter, you should not input it to the network. Feature selection, the process of finding and selecting the most useful features in a dataset, is a crucial step of the machine learning pipeline. Unnecessary features decrease training speed, decrease model interpretability, and, most importantly, decrease ...


0

VGG is a more basic architecture which uses no residual blocks. Reset usually perform better then VGG due to it's more layers and residual approach. Given that resnet-50 can get 99% accuracy on MNIST and 98.7% accuracy on CIFAR-10, it probably should achieve better than VGG network. Also, the validation accuracy should not be 100%. You could try increasing ...


0

After finding the paper authors' Github, I saw that, although they only have a MobileNet V2 model implemented, they choose the Subsample-after-ReLu option (the first one in the question). Although this doesn't fully answer my question, I'll take "the paper authors do it this way" as enough reason to prefer this over the alternative.


4

A more efficient way would be creating a multi input model, with something like this: ___________ _____________ |__Image__| |Other input| _____|_____ _____|_____ |___CNN___| |__Dense___| _____|______ _____|______ |_Features1_| |_Features2_| __|_____|__ |__Merge___| _____|______ |___Dense__| ...


1

You can find leaderboards as well as code at this address. For now, HRNetV2 leads the game. The U-Net architecture is part of a broad family of network architectures that aggregate multi-scale features to extract finer details useful for semantic segmentation. Examples are Feature Pyramidal Networks (FPN), Hourglass, Encoder-Decoder, MatrixNet, etc...


0

In the case of applying both to natural language, CNN's are good at extracting local and position-invariant features but it does not capture long range semantic dependencies. It just consider local key-phrasses. So when the result is determined by the entire sentence or a long-range semantic dependency CNN is not effective as shown in this paper where the ...


2

You get what we call high-level features, which are basically abstract representations of the parts that carry information in the image you want to classify. Imagine you want to classify a car. The image you feed your network could be a car on a road with a driver and trees and clouds, etc. The network, however, if you've trained it to recognize cars, will ...


3

U-Net and U-Net inspired architectures have been quite popular in the medical image-related tasks ever since it was first introduced. There have been several improved versions of U-Net designed for specific tasks that followed. One such example is Attention U-Net, extremely popular for Pancreas Segmentation. Other examples of architectures that have ...


1

A Max Pool layer don't have any trainable weights. Only hyperparameters is present and they are non-trainable. The max pooling process calculates the maximum value of the filter, which consists of no weights and biases. It is purely a way to down scale the data to a smaller dimension. Hope this helps you and have a nice day!


0

Feature extraction is a way that people use pretrained model to extract information from input data. For example, image segmentation task may use the VGG network or other image classifying network for feature extraction. The output of the last convolution layer is taken. Then, the features are feed into the untrained network to get outputs. The bottom ...


1

This is probably not going to work well as a way to make money. People with far larger budgets, and far more training, are already milking out any money to be made this way. This is probably their day job, and they are good at it. That said, here are some ideas: You do not need or want to use a convolutional network for this. Convolutional networks are ...


1

Mixing loss functions is very possible. For example, in the case of neural style transfer, there is a style loss and a content loss. Both of them are backpropagated through the network. The final loss used for the backpropagation is a weighted sum of the losses. In the case of style transfer, it ensures that the image generated is not only imitating the ...


Top 50 recent answers are included