RE: Give me somthing To code 04-23-2016, 07:54 PM
#13
I might be able to help you somewhat. I have an idea of a project which I am sharing with you. I will be very interesting and a little challenging, if the curiosity bug and hunger for more is there within you then feel free to work on it. Actually anybody on this forum can work on it. You will learn loads that much I can tell for sure.
Project Idea
Suppose you have a dataset of 10k images and their description/caption is also given, the text is not more than 5-10 words maybe, explaining the content of image. You are asked to use Twitter Streaming API and collect tweets for some specific hashtags, say #motivation, #inspiration, or #sad. Please note that all the tweets should contain text only and no images. Now based on the type/sentiment/meaning of tweet you have to select the most suitable picture/s'.
Example
Say you crawled a quote: “Life is not always perfect. Like a road, it has many bends, ups and down, but that’s its beauty.” ― Amit Ray, World Peace: The Voice of a Mountain Bird
From the quote you take keywords: Life, like a road, bends, ups and downs, beauty.
Now you have an image with the caption, "A beautiful lonely road with bends, ups and downs."
![[Image: lonely-road-antonia-citrino.jpg]](http://images.fineartamerica.com/images-medium-large-5/lonely-road-antonia-citrino.jpg)
Now if you get a sentiment score for the quote and also for the caption of the image then you will find that it should be somewhat close to one another. You can even analyse them with keywords too or both keywords and sentiment score. So whenever your system analyses a given quote a sample corresponding image will be shown.
Get a dataset of images, simply look it up on Google, there are many free public datasets available or you can crawl and create your own dataset. I just gave you an idea what kind text you wanna work on you can decided the main idea remains unchanged. Even if you find this challenging to code, still you can design a Software model for this as well as design the Algorithm too, steps to process data, crawl and clean data etc.
Get creative and start hacking
If you need references feel free to ask. NOTE: I am not very active on the forum but I do visit SL everyday so don't think you will not get a reply
Project Idea
Suppose you have a dataset of 10k images and their description/caption is also given, the text is not more than 5-10 words maybe, explaining the content of image. You are asked to use Twitter Streaming API and collect tweets for some specific hashtags, say #motivation, #inspiration, or #sad. Please note that all the tweets should contain text only and no images. Now based on the type/sentiment/meaning of tweet you have to select the most suitable picture/s'.
Example
Say you crawled a quote: “Life is not always perfect. Like a road, it has many bends, ups and down, but that’s its beauty.” ― Amit Ray, World Peace: The Voice of a Mountain Bird
From the quote you take keywords: Life, like a road, bends, ups and downs, beauty.
Now you have an image with the caption, "A beautiful lonely road with bends, ups and downs."
![[Image: lonely-road-antonia-citrino.jpg]](http://images.fineartamerica.com/images-medium-large-5/lonely-road-antonia-citrino.jpg)
Now if you get a sentiment score for the quote and also for the caption of the image then you will find that it should be somewhat close to one another. You can even analyse them with keywords too or both keywords and sentiment score. So whenever your system analyses a given quote a sample corresponding image will be shown.
Get a dataset of images, simply look it up on Google, there are many free public datasets available or you can crawl and create your own dataset. I just gave you an idea what kind text you wanna work on you can decided the main idea remains unchanged. Even if you find this challenging to code, still you can design a Software model for this as well as design the Algorithm too, steps to process data, crawl and clean data etc.
Get creative and start hacking
If you need references feel free to ask. NOTE: I am not very active on the forum but I do visit SL everyday so don't think you will not get a reply
![[Image: OilyCostlyEwe.gif]](http://fat.gfycat.com/OilyCostlyEwe.gif)
![[+]](https://sinister.ly/images/modern/collapse_collapsed.png)