Machine Learning In Nigeria, Today ?

The 2017 Women Techmakers Summit in Lagos, which was one of 14 other global summits, but the only one in Africa, successfully held at the Lagos Business School in April.

It was colorful, engaging, and filled with many moments I believe the 300+ attendees are yet to forget.

Themed, “Telling Your Story”, it featured amazing talks and sessions by a mix of rockstar speakers from Google, Speechless, Andela, SheLeadsAfrica, and many notable women in tech organizations / initiatives from around Nigeria.

I like that there were two workshops in the afternoon. One by Speechless on public speaking, tagged “Develop Your Story”, and another on Tensorflow — Google’s open source Machine Learning API.

Of course, both workshops were a hit ...

It Was My Last Engagement At Google Developer Relations As Community Manager For SSA

In the buildup to the event and as I prepared to co-facilitate the Tensorflow workshop, I kept thinking of how one can make Machine Learning (ML for short) valuable to African businesses and startups today, and if there are core use cases I could develop that would make ML more relatable to the attendees.

Sure, many of the great features we enjoy and take for granted in some of today’s most compelling apps are ML driven; like good spam filtering and smart responses in GMail, automatic image recognition and categorization in Google Photos, or even face recognition for easy tagging in images uploaded to Facebook .

The question I pondered on mostly was, how can businesses in Africa enable cool features and services or create better user experiences for existing ones, using machine learning ?

I decided I was going to cite the usual suspects — e-commerce and entertainment, as verticals in which Africans can derive many quick wins and great buy-in from ML. I shared my thoughts with the instructor and she brought them to live as great ML examples during the course of the Tensorflow workshop.

All went well - except I never published this article as planned. I left it in draft since then and it’s been almost 3 months now!

Prompt From The Recent Lagos Flood

This weekend wasn’t going to pass without this article getting out. Whilst thinking about it, I could not help but remember the daily grind, struggles, and realities of many Lagos residents — mad traffic, congestion, indiscriminate refuse disposal, reckless driving and parking, expensive rent requiring 1 year (or more) upfront bulk payment, touts (or louts if you prefer), bad roads, poorly maintained drains, flooding, e.t.c.

This year I experienced the annual Lagos floods first hand. There was a time the water got up to just below my belt. Don’t take my word for it, see some photos below ..

Can Machine Learning solve these in Nigeria today? This is worth exploring, and I am glad I am not the only one thinking in this direction

The recent Lagos flood made me think of Venice .. even a Lagos designed and built like Venice. Then I suddenly remembered that Machine Learning can help me visualize my flooded Lagos through the eyes of Venice.

Using I transformed some photos of the recent Lagos flood with a painting of Venice. A Venice-themed but canoe-less Lagos might look thus :

Someone has also been busy with crazy random thoughts for Machine Learning in Nigeria.

Joke Apart, Machine Learning To The Rescue

Ok. Besides the novelty of a Venice-themed flooded Lagos, and algorithms to optimize self driving cars, also for an annually flooded Lagos- what might we benefit from Machine Learning in Africa today ?

Image Recognition & Classification

Many African apps that are heavy on images can improve features and user experience with image recognition, classification and tagging.

A search on Konga for “smart phone” only reveals items with ‘smart phone’ (including the space) in the name. Same with “smartphone”. And none of the results include the iPhone, a smartphone (or smart phone)that Konga has in its catalog ?

An e-commerce platform can optimize this user experience by automatically tagging items in their catalog with metadata from the item’s image. The Google Cloud Vision API can automatically identify an iPhone as a smartphone just by analyzing a photo of it. This can then be used to tag the item and build an index for identifying it during a search.

I’ll also love to search for only hotels with a swimming pool. What about hotels with a golf pitch. Analyzing their image bank can reveal these.

Video Tagging, Labelling & Cataloging

I am a dad and my son just turned a year. We have bonded so well that my wife sometimes gets jealous. He enjoys cartoons and loves to dance when the cartoon characters are either singing or dancing. He does same with movies. How can iROKOtv help me surface Nigerian movies with singing or dancing scenes?

I have a Kenyan friend who visited Lagos for the first time and got the scare of her life when a car she was in got stopped by gun-wielding Nigerian policemen. Though relatively normal to many Nigerians, it was her first time in close range with such firearms. Before her next visit, can iROKOtv help her get more comfortable with the Police realities in Nigeria by surfacing Nigerian movies with Police or gun scenes?

A search on iROKOtv for “Police” or “Traffic” reveals nothing, and I think this is because there are no movies or actors / actresses with these words in their names.

For a 2018 valentine ad campaign, how will iROKOtv create a highlight video of the best kiss scenes from its catalog? If iROKOtv has The Wedding Party in its catalog, will a search for “Dance” surface it because it has (IMHO)one of the best dance scenes in a Nigerian movie?

The Cloud Video Intelligence API from Google can help address these use cases. Like the Vision API, it analyzes videos and surfaces metadata that can be used to identify, categorize, and search them.

Call To Action

There is huge potential for what Machine Learning can do in Africa, today! The big tech companies (Google, Apple, Microsoft e.t.c) have named Artificial Intelligence the future — not apps, not mobile, not the operating system — but the ability for us to empower computers to learn, do and be, and the potential it holds. This is the ability to enable transformative experiences.

With tools like Tensorflow, Cloud Vision API and Video Intelligence API, we don’t have to wait till we can implement algorithms to make self driving cars work in our terrain. We can begin to get value right away

Our users are getting great experiences using Google Photos, Facebook, e.t.c and have high expectations. Besides, Sundar Pichai is watching ..

I’ll like to hear from you. Do share your thoughts in the comments section below or tweet @chaluwa with any questions or suggestions. Also, don’t forget to recommend and share this article if you found it useful.