Some 200-300 event photos could be generated for a typical university event. Out of all of the photos, there are fuzzy ones, underexposed ones and the ones where everybody seems unhappy for some reason - it takes a huge amount of time and efforts to handpick the useable photos from the pool for social media use.
Pic It! is an Android app that helps event organizers efficiently filter and rank the best, most relevant photos from all photos.
Current solutions on the market
Most of the photo processing applications focus on the individual photo rather than comparing across multiple photos. Out of the ones who do compare between many input images, like Google Photo, EyeEm Select, or The Roll, none of them is customized for event/group photos. Usually, the algorithm for "good" photos is too generic to produce a useful outcome.
What Pic It! does
Pic It! Ranks photos using our algorithm that takes into account: (1) the quality of the photo; (2) the average positive emotions on every detectable face in the frame that was analyzed using machine learning tools.
Pic It! Filter photos by how relevant a tag/theme is to the photo.
Pic It! and Women
Pic It! allows the user to choose photos with better gender ratios to feature the female representation in the collection.
Seeing more women participating in events will set examples for others, making women themselves more confident and more involved in fields where they are traditionally less populated - whether be it starting a business, or taking part in STEM education. Our algorithm also has the capability of taking into account the gender ratio of the image to help rank photos that help promote diversity more!
How we built it
We developed a mobile application in Android Studio, employed Machine Learning algorithms powered by Google Vision API / MLKit and Microsoft Face technologies, and implemented image labeling as well as face and gender recognition.
By combining both APIs, we created our own algorithm that calculates a score for each photo, and constantly updates the score when a different tag/theme is selected.