Building Convolutional Neural Networks with Tensorflow

Building Convolutional Neural Networks with #Tensorflow #abdsc

  • In the past year I have also worked with Deep Learning techniques, and I would like to share with you how to make and train a Convolutional Neural Network from scratch, using tensorflow.
  • Later on we can use this knowledge as a building block to make interesting Deep Learning applications.
  • Source code is also provided.
  • Before you continue, make sure you understand how a convolutional neural network works.
  • The code is also available in my GitHub repository, so feel free to use it on your own dataset(s).

In the past year I have also worked with Deep Learning techniques, and I would like to share with you how to make and train a Convolutional Neural Network from…
Continue reading “Building Convolutional Neural Networks with Tensorflow”

[1609.03677v1] Unsupervised Monocular Depth Estimation with Left-Right Consistency

Unsupervised Monocular Depth Estimation. Awesome red-eye read!  #deeplearning #depth #stereo

  • Abstract: Learning based methods have shown very promising results for the task of depth estimation in single images.
  • By exploiting epipolar geometry constraints, we generate disparity images by training our networks with an image reconstruction loss.
  • Most existing approaches treat depth prediction as a supervised regression problem and as a result, require vast quantities of corresponding ground truth depth data for training.
  • Our method produces state of the art results for monocular depth estimation on the KITTI driving dataset, even outperforming supervised methods that have been trained with ground truth depth.
  • We propose a novel training objective that enables our convolutional neural network to learn to perform single image depth estimation, despite the absence of ground truth depth data.

Continue reading “[1609.03677v1] Unsupervised Monocular Depth Estimation with Left-Right Consistency”

models/lm_1b at master · tensorflow/models · GitHub

Language Model trained on 1 billion words, 800k vocabulary size. Code in TensorFlow.

  • (omitted) Eval Step: 4529, Average Perplexity: 29.243668.
  • (omitted some TensorFlow output) Finished softmax weights Finished word embedding 0/793471 Finished word embedding 1/793471 Finished word embedding 2/793471 …
  • Given provided dataset, calculate the model’s perplexity.
  • (omitted) # Run eval mode: bazel-bin/lm_1b/lm_1b_eval –mode eval \ –pbtxt data/graph-2016-09-10.pbtxt \ –vocab_file data/vocab-2016-09-10.txt \ –input_data data/news.en.heldout-00000-of-00050 \ –ckpt ‘ data/ckpt-* ‘ …
  • # Run dump_lstm_emb mode: bazel-bin/lm_1b/lm_1b_eval –mode dump_lstm_emb \ –pbtxt data/graph-2016-09-10.pbtxt \ –vocab_file data/vocab-2016-09-10.txt \ –ckpt ‘ data/ckpt-* ‘ \ –sentence ” I love who I am . “

models – Models built with TensorFlow
Continue reading “models/lm_1b at master · tensorflow/models · GitHub”

Building a Data Science Portfolio: Machine Learning Project Part 1

Building a #DataScience Portfolio: Machine Learning Project Part 1  @dataquestio

  • KDnuggets Home > News > 2016 > Jul > Tutorials, Overviews > Building a Data Science Portfolio: Machine Learning Project Part 1 ( 16:n27 )
  • Dataquest’s founder has put together a fantastic resource on building a data science portfolio.
  • A loan that is acquired may have dozens of rows in the performance data.
  • A good way to think of this is that the acquisition data tells you that Fannie Mae now controls the loan, and the performance data contains a series of status updates on the loan.
  • A good dataset for an end to end portfolio project can be hard to find.

Read the full article, click here.


@kdnuggets: “Building a #DataScience Portfolio: Machine Learning Project Part 1 @dataquestio”


Dataquest’s founder has put together a fantastic resource on building a data science portfolio. This first of three parts lays the groundwork, with subsequent posts over the following 2 days. Very comprehensive!


Building a Data Science Portfolio: Machine Learning Project Part 1

Demystifying Machine Learning Part 4: Image and Video Applications

Demystifying #MachineLearning Part 4: Image and Video Applications from @AnandSRao:  #AI

  • Deep learning requires large volumes of data in order for a system to learn the features and should not be attempted where data is sparse.
  • , deep learning is suitable only for certain classes of problems and cannot be seen as a panacea to solve all problems.
  • In the previous post in our Machine Learning series, we dived into the inner workings of deep learning .
  • Previous post: Sportifying STEM through Robotics to Stimulate Learning
  • Demystifying Machine Learning Part 4: Image and Video Applications

Read the full article, click here.


@PwCAdvisory: “Demystifying #MachineLearning Part 4: Image and Video Applications from @AnandSRao: #AI”


How are companies using deep learning to drive business goals? Improved image and video recognition, audio recognition, and language understanding.


Demystifying Machine Learning Part 4: Image and Video Applications

Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic Image Colorization with Simultaneous Classification

End-to-end #MachineLearning: Automatic Image Colorization with Classification  #DataScience

  • Based on Convolutional Neural Networks, our deep network features a fusion layer that allows us to elegantly merge local information dependent on small image patches with global priors computed using the entire image.
  • We present a novel technique to automatically colorize grayscale images that combines both global priors and local image features.
  • The entire framework, including the global and local priors as well as the colorization model, is trained in an end-to-end fashion.
  • The output of our model is the chrominance of the image which is fused with the luminance to form the output image.
  • Our architecture can process images of any resolution, unlike most existing approaches based on CNN. We leverage an existing large-scale scene classification database to train our model, exploiting the class labels of the dataset to more efficiently and discriminatively learn the global priors.

Read the full article, click here.


@MikeTamir: “End-to-end #MachineLearning: Automatic Image Colorization with Classification #DataScience”


We present a novel technique to automatically colorize
grayscale images that combines both global priors and local
image features. Based on Convolutional Neural Networks, our
deep network features a fusion layer that allows us to
elegantly merge local information dependent on small image
patches with global priors computed using the entire image. The
entire framework, including the global and local priors as well
as the colorization model, is trained in an end-to-end fashion.
Furthermore, our architecture can process images of any
resolution, unlike most existing approaches based on CNN. We
leverage an existing large-scale scene classification database
to train our model, exploiting the class labels of the dataset
to more efficiently and discriminatively learn the global
priors. We validate our approach with a user study and compare
against the state of the art, where we show significant
improvements. Furthermore, we demonstrate our method
extensively on many different types of images, including
black-and-white photography from over a hundred years ago, and
show realistic colorizations.


Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic Image Colorization with Simultaneous Classification