Would you interpret this image as a correct deconvolution process?

I’m applying the function conv2d_grad_wrt_inputs in theano to deconv a feature map into the original image. In the figure below the first image to the left is the input image which I’m applying convolution and the result is the feature map in the second image while the third image is the deconvolution process. I’m applying … Read more

How would you interpret decreasing cost but increasing training and validation error during epochs?

I’m training a fully convolution network with final layer of global sum pooling and no intermediate pooling in the network which I already test on global average pooling and it converged very well. The reason I’m testing the global sum is to penalize harder the activation map and make them only sensitive to the actual … Read more

3d Convolution vs CNN-LSTM for Gesture recognition

I want to implement a gesture recognition system from video (of hand movements). Some people have experimented with 3d convolutions to extract not only spatial features out of images, but also extract temporal features. My question is, how does this architecture compare with a CNN on a per frame basis, to an LSTM network ? … Read more

Understanding the second hidden layers of convnets

Having studied ordinary fully connected ANNs, I am starting to study convnets. I am struggling to understand how hidden layers connect. I do understand how the input matrix forward feeds a smaller field of values to the feature maps in the first hidden layer, by moving the local receptive field along one each time and … Read more

Discrepancy between categorical cross entropy and classification accuracy

I have a convolution neural network with random weights initialized and Trained to perform binary classification. I have 2000 images as training data and 2000 validation data. The problem I am trying to solve is if the image is healthy or not.The loss function used is categorical cross entropy. I have implemented the model in … Read more

How to overcome overfitting? [closed]

Closed. This question needs to be more focused. It is not currently accepting answers. Want to improve this question? Update the question so it focuses on one problem only by editing this post. Closed 3 years ago. Improve this question I am using MatConvNet for the classification two different pants. i am using MINST example … Read more

Different sized inputs for batch training in fully CNNs

The idea of transfer learning is to use already trained networks for another purposes to the one it was initially trained for. Using fully convolutional networks, the activation maps can be used to create descriptors that later on can be used to train a classifier for image classification. I am currently using a pretrained VGG … Read more

Precise localization and characterization of rudimentary shapes with neural networks

I understand that there are flavors of (convolutional) neural networks that are useful for object localization and detection tasks of reasonable difficulty. In all of the examples I have seen so far, localization is formulated as finding the corners of a bounding box. Often, the fit is not required to be very precise: Conversely, I … Read more

How to retrain a model (Inception) with ‘prioritised’ images in certain classifications

I am new to machine learning, and have constructed a basic CNN classifier by retraining the last layer of the Inception v3 model with my own image set into two classifications. I did this in Python using Tensorflow, following the guidelines from here. I used two files to achieve this: retrain.py : Retrains the Inception … Read more

Location Invariance not achieved in CNN in spite of 99% test accuracy

I have been trying to decode a captcha using CNN. The number of training samples is 1,82,000. So far, I could achieve 99% accuracy on training as well as test set. The issue however is that LOCATION INVARIANCE is not achieved while training. The captcha text if slightly shifted horizontally gives bad prediction. Below image … Read more