Report Outline, Presentation and Deadline

[mbarbera]

I attach here the link for the online word file, once you open the link click "edit document" and then "edit in the browser" to access the edit version. The report is due on 14th of November by 11.59 pm. Remember that the presentation is scheduled either on the 26th of October or on the 2nd of November.

FYI: The exam is scheduled on the 22nd of November from 1:30 PM to 4:30 PM

Here is the link:

https://entuedu-my.sharepoint.com/personal/n1609543k_e_ntu_edu_sg/_layouts/15/guestaccess.aspx?guestaccesstoken=AF%2bCKVMi0XPAZKFFALxlPfC8Xr1K%2fXu%2bXzUJgWpoNLo%3d&docid=1606d5db4fcee48eba2ee2e630a605228&rev=1

[mbarbera]

This is the link for the presentation. Right now is blank, but I will start to fill it up with the outline. You're more than welcomed to contribute

https://entuedu-my.sharepoint.com/personal/n1609543k_e_ntu_edu_sg/_layouts/15/guestaccess.aspx?guestaccesstoken=2V1yG7lt4vBj0naI9t6F%2fuUWgvsRUKo9RVHuhYOM5Ww%3d&docid=140d0933300e344d9ae31066654deb1e9&rev=1

[JanEr93]

added some remarks to the presentation, forgot to sign in so I did them with the guest user.

[sacdallago]

I'm adding stuff to the presentation. Note: Not everything I put might be necessary, but it's better to have more and to cut than the other way around

[JanEr93]

Report format: 12 point font, single space, and the maximal number of pages of the project report is 20 excluding reference

[JanEr93]

@mbarbera @ripitrust I think we should add two/ three more slides to the presentation and remove two:

• 1 for the feature creation which was not successfull (with examples of skeletonize, binarization ...)
• 1 or 2 for tensorflow and convolutional neural network including a rough outline of its function.
• remove the slides on the team and our project organization (meetings...)

We should also use this information in our presentation (maybe when presenting the labels (I did this already and include it into the presentation):

@mbarbera We should include the histogram in the report as well.

@sacdallago The histogram shows the frequency of the labels in the (original) business2labels and in our images2labels csv-files. Do you think we can make use of this information (overestimating/ underestimating the frequency)? E.g. by adjusting the threshold for those levels (I was also thinking about setting label_dependet thresholds (meaning a vector of 9 thresholds).

[sacdallago]

I agree on the slides!

For the proportions in training, look at my latest committed file: https://github.com/sacdallago/dataminer/blob/master/conv_net_one_against_all.ipynb

We need to store the model now and perform predictions for the test images

[sacdallago]

Whops, closed by mistake

[JanEr93]

@sacdallago Yes, that is where I got the label frequency among the images from (blue). The orange histogram is the 'original' distribution of the labels, before assigning the labels to the pictures directly. So, as you can see, for some labels we have a higher/lower proportion than in the original business2labels file.

You are right about the storing of the model... Did you make any progress here, @ripitrust ? I will set up a piece of code retrieving the business to labels mapping from the pic2image and pic2biz.

[ripitrust]

@JanEr93 I am editing the slides

how would you like these to be added ?

[mbarbera]

UPDATE: I am currently writing the report and I would say I am missing just small parts. In Particular the Results part and the Conclusion. I would be great if you guys can provide me an outline with the summary of the Results. I also pointed out in the report the Group member information and activity level, if you don't agree on the description or you want to add/delete something let me know :)

Here's the file: Group14_report.pdf

[sacdallago]

@mbarbera working on that :D

[ripitrust]

@mbarbera I will write an introduction about TensorFlow

[mbarbera]

@ripitrust I already wrote about TensorFlow. Try to integrate if you want my section 👍

[ripitrust]

@mbarbera great I will help integrate it

[mbarbera]

Updated report:

Report.pdf

[mbarbera]

Last version of the Report:

Group14_Report.pdf

[ripitrust]

@mbarbera Done with the TensorFlow Architecture Architecture of TensorFlow.docx

[ripitrust]

References in APA Style

Shukla, N. (2016). Machine Learning with TensorFlow (MEAP ed.). Manning.

TensorFlow API Documentation. Retrieved November 12, 2016, from https://www.tensorflow.org/versions/r0.11/api_docs/index.html

[sacdallago]

@ripitrust that's IEEE 😆

[ripitrust]

@sacdallago ohh I should remove the numbering in front

[mbarbera]

Last version of the Report:

Group14_Report.pdf

[mbarbera]

Last version of the Report:

Group14_Report.pdf

As far as I am concerned we can actually upload it, if you guys want to edit it a little bit, here is the link of sharelatex. Maybe you want to add something on the Results or Conclusion part

https://www.sharelatex.com/project/581da586d867d54711ef15e9

[mbarbera]

Last version of the Report, GitHub Repo cited

Group14_Report.pdf

[sacdallago]

I'd say: done :)