Loading... 0%

Nov 18, 2019

Hackfest for Haribon & other efforts

By Senti AI

undefined
Raj Mariano (leftmost) with the ML Hackfest Winners (left to right): Jero Santos, Nam Desembra, Nyessa Michelle Kho, Jude Teves, Prince Joseph Erneszer Javier, Marc Reyes.

Hosted by Senti AI, ML Hackfest 2019 (Machine Learning hackfest) is a hackathon that aims to encourage the local Data Science and Artificial Intelligence communities to test their skills and use them to help NGOs overcome data-related challenges that prevent them from making the Philippines a better place.

The event is centered around "machine learning" - a subset of artificial intelligence where systems (or models) have the ability to automatically learn and improve from experience without human help. This capability can help organizations like NGOs overcome data-related challenges due to lack of manpower and expertise.

"We really believe in the purpose of this hackathon so we want to make sure that the Machine Learning models created by the participants will not go to waste and will actually help the NGO partners." says Ralph Regalado, CEO of Senti AI and Google Developer Expert for Machine Learning.

Took place last November 18 at Google Philippines, the ML Hackfest 2019 is supported by the biggest AI community in the Philippines, AI Pilipinas, along with Senti AI, the Philippines' pioneering Artificial Intelligence company as part of the Philippine Startup Week 2019. 

The event allowed students, Data Scientists, Machine Learning Engineers, and AI enthusiasts to collaborate and create solutions that address an NGO partner's specific problem within a defined time frame.

The winning teams were judged based on their Model Performance: "F1-Score" (a measure of a test's accuracy) for the Haribon Challenge and "Intersection Over Union" (the most common metric for an Image Segmentation Problems) for the DOST-DATOS challenge. In the case of a tie for the model performance, it will be based on who has the lower number of "trainable parameters" for the Machine Learning model through a code review.

There were two winning teams - one for each challenge. The first challenge was the Haribon Foundation Challenge, where participants must accurately classify the most suitable species for a given location feature. Team Hanash, whose members are Jero Santos, Nam Desembra, and Nyssa Michelle Kho, won this challenge. 

According to an interview with Senti AI's Research Engineer, Justine Clemente, Team Hanash used Auto ML (a platform where you can build machine learning models automatically) to generate their algorithm in classifying the species. It was a surprising turn of events that Team Hanash won the Haribon challenge despite lagging at 7th place during the preliminary scoring. However, upon comparing Team Hanash's predictions to a more complete data set after the contest (which is the deciding factor), it was revealed that the group's work was the most accurate.

"We tried different kinds of algorithms. We even tried to grid search or random search, but we kept getting low scores. In the end, we tried to tune the Auto ML, where we finally got the high score," said Nyssa Kho of Team Hanash. 

Apart from Haribon, DOST-DATOS also contributed a second and more difficult challenge for the participants. Many teams tried the DOST-DATOS challenge which almost ended up in a tie between Team Hackdog from Novare Technologies and Team Gwapings.

Ultimately, a decision was achieved through a careful review of the codes. While Team Hackdog created a model with 24 million trainable parameters, Team Gwapings composed of Jude Teves (from Asian Development Bank), Prince Javier, and Marc Reyes (from Mynt), achieved the same results with ONLY 2 million trainable parameters. This showed that Team Gwaping's model is more optimized and, thus, securing their place as winners of the DOST DATOS challenge.

The DOST-DATOS Challenge is a semantic segmentation challenge where the participants must create a Machine Learning model that can automatically identify bare land, agricultural areas, vegetation, vegetation, water, urban areas, clouds, and haze from satellite images of the Philippines. The goal is to create a Machine Learning model with the highest accuracy in terms of classifying each pixel of the image.

"It's best to start simple. If I'm going to make a neural network, I don't want to put ten layers immediately. Start simple, if two layers can work, then you want to improve it by adding more layers, then add layers incrementally. That's how we did it. We thought 'let's try a very simple network, but we also tried a very complicated network, which had 64 million parameters, so we tried both, and the simpler network was easier to train and also had better accuracy," according to Prince Javier, from Team Gwapings. 

One of the organizations partnered with ML Hackfest is Ambidextra content marketing, and events management company focused on serving tech companies in the Asia Pacific. 

"Ambidextr fully supports the growth of the technology industry in the Philippines. As a full-stack marketing agency helping tech companies communicate the positive work that they do, we are happy to be partnering with Senti AI for this year's ML Hackfest," said Robby Vaflor, writer at Ambidextr.

All the winners got to take home cash, gadgets from Accenture in the Philippines and PR exposure. While all the participants of the hackathon were treated to gadgets, giveaways, and dinner.

This event would not be possible if it weren't for all the sponsors and partners:

ABOUT SENTI AI
Senti Techlabs Inc. (also known as Senti AI) is the pioneering Artificial Intelligence company in the Philippines. The company's focus is to continuously create algorithms that would make computers understand and adapt to the way that Filipinos think and communicate.

ABOUT AI PILIPINAS
A group for Filipinos who are willing to upskill tech capabilities and learn Machine Learning/Artificial Intelligence. These skills can either be used for research purposes and/or developing end-to-end solutions. All skill levels are welcome

ABOUT ACCENTURE IN THE PHILIPPINES
Accenture is a leading global professional services company, providing a broad range of services and solutions in strategy, consulting, digital, technology and operations. Combining unmatched experience and specialized skills across more than 40 industries and all business functions - underpinned by the world's largest delivery network - Accenture works at the intersection of business and technology to help clients improve their performance and create sustainable value for their stakeholders. With 492,000 people serving clients in more than 120 countries, Accenture drives innovation to improve the way the world works and lives. Visit us at www.accenture.com

ABOUT ANC
ANC (ABS-CBN News Channel) is the Philippines' first 24-hour news channel and the only cable news network running 24/7. The channel features news updates and breaking news from the Philippines and around the globe.

Help us protect
more communities

Array
(
    [0] => stdClass Object
        (
            [id] => 15
            [form] => 
            [form_type] => 
            [type] => Amount
            [value] => 150
            [parent_id] => 0
            [sub_type] => 
            [sub_value] => 
            [sub_cat_img] => 
            [sub_cat_age] => 
            [no_of_seedling] => 0
            [amount] => 0.00
            [description] => 
            [order_number] => 
            [created] => 2023-01-30 14:15:10
            [modified] => 2023-04-13 14:13:10
            [author_id] => 1
            [is_deleted] => 0
        )

    [1] => stdClass Object
        (
            [id] => 16
            [form] => 
            [form_type] => 
            [type] => Amount
            [value] => 500
            [parent_id] => 0
            [sub_type] => 
            [sub_value] => 
            [sub_cat_img] => 
            [sub_cat_age] => 
            [no_of_seedling] => 0
            [amount] => 0.00
            [description] => 
            [order_number] => 
            [created] => 2023-01-30 14:15:15
            [modified] => 2023-09-15 16:39:48
            [author_id] => 1
            [is_deleted] => 0
        )

    [2] => stdClass Object
        (
            [id] => 17
            [form] => 
            [form_type] => 
            [type] => Amount
            [value] => 1000
            [parent_id] => 0
            [sub_type] => 
            [sub_value] => 
            [sub_cat_img] => 
            [sub_cat_age] => 
            [no_of_seedling] => 0
            [amount] => 0.00
            [description] => 
            [order_number] => 
            [created] => 2023-01-30 14:15:22
            [modified] => 2023-09-15 16:39:59
            [author_id] => 1
            [is_deleted] => 0
        )

    [3] => stdClass Object
        (
            [id] => 18
            [form] => 
            [form_type] => 
            [type] => Amount
            [value] => 5000
            [parent_id] => 0
            [sub_type] => 
            [sub_value] => 
            [sub_cat_img] => 
            [sub_cat_age] => 
            [no_of_seedling] => 0
            [amount] => 0.00
            [description] => 
            [order_number] => 
            [created] => 2023-01-30 14:15:22
            [modified] => 2023-09-15 16:40:09
            [author_id] => 1
            [is_deleted] => 0
        )

)