I have written previously about adapting my convolutional classifier for mammograms to a network that will do dense prediction for each pixel, resulting in the segmentation of the images into normal and abnormal.
I recently made two major changes, both of which drastically improved the results. The changes were using resizing rather than transpose convolutions to upsample and changing the prediction from a hardmax to a softmax.
When I started working adapting the classifier to do dense prediction I followed the blueprint from “Fully Convolutional Networks for Semantic Segmentation” by Long, et al. [23] Very simply this method removed the…
My original work with the DDSM and CBIS-DDSM dataset yielded good accuracy and recall on the test and validation data, but the model didn’t perform so well when applied to the MIAS images, which came from a completely different dataset. Additional analysis of the images indicated that the negative images (from the DDSM) and the positive images (from the CBIS-DDSM) were different in some subtle but important ways:
- The negative images had lower contrast.
- The negative images had a lower mean, lower maximum and higher minimum pixel value.
We had become concerned about point 1 when we discovered that increasing…
Introduction
Breast cancer is the second most common cancer in women worldwide. About 1 in 8 U.S. women (about 12.4%) will develop invasive breast cancer over the course of her lifetime. The five year survival rates for stage 0 or stage 1 breast cancers are close to 100%, but the rates go down dramatically for later stages: 93% for stage II, 72% for stage III and 22% for stage IV. Human recall for identifying lesions is estimated to be between 0.75 and 0.92 [1], which means that as many as 25% of abnormalities may initially go undetected.
The DDSM is a…
I recently set up a Google Cloud instance to train some TensorFlow models on. While Amazon EC2 has AMIs that already have everything configured for you, on Google Cloud you need to set up everything yourself. I spent several days doing this, and the instructions I found online — both from Google and from other places were for older versions of TensorFlow so did not work with the newest version, which at the time of this writing is 1.7.
These instructions will work for v1.7 and have been tested several times. I hope they will help someone from having to…
Lately I keep hearing about Facebook’s “data breach” and how Cambridge Analytica “stole” private data from Facebook. I am not a big fan of Facebook, but in this case I feel I have to come to their defense.
This was absolutely not a “data breach.” I have never installed third party apps on Facebook so I am not sure of the details, but I assume that when you install one you have to accept its conditions and grant it permission to access your data. I would hope that you would be notified of what data the app will access, like…
The response to the recent mass shooting reaffirms my decision to have left the US and moved to Switzerland. In most other first world countries, the responses to mass shootings have been to ban or restrict guns. This seems to have worked very well in places like the UK and Australia.
In the US the response to a school shooting is to have the teachers carry guns. While this probably would reduce the number of deaths in a school shooting, the thinking behind it is just so backwards it boggles my mind. …
In 2011 or so I bought $1,000 worth of Bitcoins. At the price at the time that was about 100 BTC which would today be worth about $1 million. I could easily be kicking myself for not having kept them for years but I never bought them in the hopes they would grow in value. I got them to spend them, and spend them I did, although I will avoid self-incrimination and not go into what I spent them for.
For me, Bitcoin has always been a medium of exchange. If I needed some for any reason I would buy…
Aujourd’hui, la politique américaine est tout au sujet d’argent, et les politiciens passent une quantité excessive de temps à élever l’argent. Le coût d’être élu ne cesse pas d’augmenter. En 1960 John Kennedy et Richard Nixon a dépensé $19.8 millions sur ses campagnes présidentielles. L’ajustement de ces montants pour l’inflation rend $157 millions aujourd’hui. En 1960, c’était impensable pour les candidats de dépenser autant d’argent sur une élection. En 2012, le coute de l’élection présidentielle des États-Unis était $2.6 milliards, et cette augmentation massive des coûts est due, en grand partie, à la publicité. …
When I first arrived in Switzerland I was appalled at the level of paperwork and bureaucracy required to do even the simplest thing. When I first heard about the Bilag I counted it as another piece of red tape that needed to be dealt with here. The Bilag is a CHF 400 tax you have to pay to the government in exchange for the right to own any device capable of receiving broadcast signals — a TV, a radio, even a computer. …
