training result changes of each run in TensorFlow official example - machine-learning

I am running the example of "MNIST For ML Beginners" (https://www.tensorflow.org/get_started/mnist/beginners). The official code is here: https://github.com/tensorflow/tensorflow/blob/r1.2/tensorflow/examples/tutorials/mnist/mnist_softmax.py
Then I found the results are different if I ran it multiple times. The question is how can this happen if there is no randomization in the code?

By looking at the source code here:
https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/learn/python/learn/datasets/mnist.py#181, it is actually the training examples changing every time.
For your convenience:
numpy.random.shuffle(perm0)

Related

How to warm start SNOPT in drake

Hi I was wondering how to warm start the snopt solver in pydrake?
In the documentation for SNOPT it says to use start =2. However I'm not sure how to feed that in properly and also send information from the previous solve into snopt
We haven't supported all the warm-start features in Drake yet. In Drake, you can give it an initial guess
result = Solve(prog, initial_guess)
There are other API's to set the initial guess. You can refer to our tutorial in the section "Using an initial guess". You can use the previous solution as the initial guess for the current solve.
We currently don't support warm-starting with dual variable or basis vectors yet.

how to pretrain my image using resnet50 in mask-rcnn

I am researching about mask r-cnn. I want to know how to pretrain my image(knife,sofa,baby,.....) using resnet50 in mask-rcnn. I struggle to find that in github, but I can't. Please help me anybody who know how to handle it.
Try this implementation of Mask RCNN on github here.
You can follow Mask_RCNN github repo. It has both resnet50 and resnet100 (might be wrong here). It is a beautiful implementation I would say. The base model is from FAIR (Facebook AI Research). There is a demo file which you can check before starting your work.
If it works well, you can see my answer, it will help you to train the model with your custom data. The answer is a bit long, but it lists all the steps.
Something which I personally like about this implementation is:
It is easy to setup. Won't bother you much about the dependencies. Having a python virtual environment does the wonders.
It falls back automatically from a CPU version to GPU and vice versa.
It is having good support from its developers. It is getting commits frequently.
The code is very customisable. So If you want to do some changes, it's pretty easy. Some booleans and numbers changes up and down and you are done...!!!

Style Transfer in CoreML

I would like to use style transfer (example) in CoreML. Since CoreML support converting Keras my first thought was to convert one of their samples like this one or this one but it seems there's few issue with this approach based on this thread.
How can I use Style Transfer in CoreML? any examples will help.
Edit:
Thanks for the link #twerdster, I was able to test it and it's working for me.
Additionally I found this (torch2coreml) repo by Prisma.

Getting ElliFit ellipse fitting algorithm to work

I have tried to implement the ellipse fitting algorithm descibed in the following paper: “ElliFit: An unconstrained, non-iterative, least squares
based geometric ellipse fitting method”, by Prasad, Leung, Quek. A free version can be downloaded online from http://azadproject.ir/wp-content/uploads/2014/07/2013-ElliFit-A-non-constrainednon-iterative-least-squares-based-geometric-Ellipse-Fitting-method.pdf
The authors did not provide any publicly available implementation.
I have implemented the algorithm in Mathematica, I believe I have implemented it correctly, yet it fails to correctly find the fit parameters. The PDF of the experiment can be downloaded here: http://zvrba.net/downloads/ElliFit-fail-example.pdf
Did somebody else try to implement this particular algorithm and, if yes, what is the key to get it working? Is there a "bug" in the paper? Can somebody take another look at my implementation and see whether there's a bug there?
I know it's been almost a year since this question, but it seems that the authors have now provided public source code for ElliFit, both a MATLAB version and an OpenCV version.
Both are available on the the author's homepage. In case the homepage goes offline for some reason, both source codes are shared on Google and are available here (MATLAB) and here (OpenCV).
At the time of writing, I have not personally tested their code, but am planning to use them for a project. I will post any updates here in the next few days.
EDIT:
I got around to test the code sooner than I expected. I gave the OpenCV code a try. It works pretty well, as demonstrated by the image below (ignore the "almost-closed-ellipses". It's an artifact caused by something else in my code).
As you can see, it works pretty well, most of the times. There are some failure cases too (the small ellipse on the spray bottle next to the cup).

multiple choice test mark reader - where to start?

I was assigned a project (in school) for automated multiple choice test scoring and I do not know where to start.
I think his is a kind of popular program and you already know about it. Enter an image file scanned of the answer sheet and return results.
Everything I know about computer vision is a few examples of photo editing with OpenCV. I hope you can give me a few keywords related to the problem or maybe a couple of blog articles, documents and related libraries.
Is there any free open source programs that I can refer to?
Thanks!
Edit: Add 2 example of the answer sheet (sory that I cannot find a sheet in English):
I think there are basically two steps to the problem
bring the form into a normalized position
now you know where the boxes are and can look at them by thresholding the gray values in that region.
What methods to use for step 1 depends on your actual images and how much the vary. Do you have some example images you can upload?
Also I think it is a good idea, especially if you are a beginner, to start with some simple examples and work your way up from there by adding more and more variation.

Resources