This project's main goal remains a "proof of concept" based on what I have studied through the material of the subject "Social Media Management".
"Instagram Hashtag Generator" analyzes an image and gives back some hashtags based on a prediction of binary classifiers. It is possible to use two different classifiers, a classifier based on Logistic Regression and the other one on k-nearest neighbors.
Instagram allows its users to use hashtags, a maximum of 30 hashtags per post precisely. Hashtags are used to make a post easier to be discovered by other users. Some hashtags are more popular than others because they are used independently of the context. However, some are correlated with each other; for example, a picture of two friends could be associated with the tags "people" and "friends").
The idea is to analyze some main tags and seek out the most common hashtags used with them, removing the ones known to be used to gain more likes and comments (e.g., "picoftheday").
Follow an example of the output of the program on two images.
The tags that has been analyzed are:
- nature
- sky
- summer
- sea
- friends
- food
- art
- street
- car
- building
- people
- animal
- cute
- happy
- fashion
- sunset
- architecture
- landscape
- smile
- girl
For each of them is possible to download the model (knn or lr) already trained with more than 10000 images in total.
It is present a pdf that is the relation I presented as a project, in which I explained each step of the realization of the algorithm.