Space is personal
The Philippines is an archipelago of more than 7,000 islands (depending on whether it’s high tide or low tide) and over 100 languages (depending on who you ask). But our Filipino ancestors’ fascination with space transcends both geography and communication. Our ancient names for constellations varied, “Orion” was known as “Balatik” by some and “Seretar” by others. Our relationships with planets greatly differed, in the north in Sagada we offered live chickens before funerals to Venus, and in the south in Sulu we worshipped Jupiter and prayed for the women to birth beautiful children . Space is personal for Filipinos. Before we even knew what “astronomy” was or meant, it had already been a part of our culture.
Source: Scout Magazine PH
Fast forward to 2021 and our pre-colonial ideas of space have evolved, but our wonder remains. Compared to our Southeast Asian neighbors, for the longest time, the Philippines has been slower to get into the space-faring game. However, this all changed when President Rodrigo Duterte signed Republic Act No. 11363 in 2019, thereby creating both the Philippine Space Agency and a Philippine Space Development Fund.
Space is also universal
While the Philippines has a lot of catching up to do with the rest of the world’s space-faring nations, this does not hinder us from celebrating mankind’s latest milestones this year, namely: Russia’s Hope mission and China’s Tianwen-1 mission entered Mars’s orbit, and NASA’s Perseverance rover
landed on Mars’s surface all in February 2021.
All this attention on Mars makes sense. It’s our next-door neighbor and the planet we are exploring for future human settlements—NASA dreams of seeing man on Mars in the 2030s (although COVID might slow down plans). Elon Musk also thinks SpaceX can beat NASA to Mars by landing humans by 2026.
We’ve had our sights set on the “Red Planet” since the 1960s and we’ve accumulated a lot of data in that time—to be specific, we’ve collected more than 300 terabits of it. Just to give you an idea, 1 terabit is equal to 125,000 megabytes.
We can credit this data set we’re talking about to the Mars Reconnaissance Orbiter (MRO), which has been in orbit around Mars since 2006, to find evidence of water that can sustain life. This is the spacecraft responsible for making 361 terabits of data available to the data scientists, data engineers, machine learning engineers, researchers, and AI and domain experts of Omdena’s “Anomaly Detection on Mars Using Deep Learning” project.
Mars is our future
Eskwelabs is excited to partner with Omdena, a collaborative platform that builds ethical and efficient AI solutions by building “cross-functional teams of 40 to 60 collaborators,” representing up to about 25 countries per project. This organization’s entrance into humanity’s space exploration narrative couldn’t have come at a better time. Their “Anomaly Detection on Mars Using Deep Learning” project focuses on finding technosignatures, which are known as “measurable properties that provide scientific evidence of past or present extraterrestrial technology.” Omdena’s project goal syncs perfectly with NASA’s life-finding mission on Mars via their Perseverance rover. Among its four main goals, Perseverance’s priority is to determine if Mars ever supported life.
So, to answer the age-old question, “Are we really alone in the universe?,” here is what the project team of Omdena did using data from NASA’s Frontier Development Lab—which Eskwelabs Data Club members will have access to try it themselves.
From Mars to Jupyter
Remember the Mars Reconnaissance Orbiter (MRO)? The spacecraft that collected 361 terabits of data? NASA built this spacecraft and gave it the ability to capture high-resolution photos of Mars's surface using its High Resolution Imaging Science Experiment (HiRISE) camera. Capturing these images was aimed at helping scientists determine potential landing sites in the future.
The MRO relays the photos it has captured back to Earth via NASA’s Deep Space Network (DSN), “which is a worldwide network of U.S. spacecraft communication facilities.” Members of Omdena’s project team then accessed this data via the Mars Orbital Data Explorer (ODE) website on their browsers, or more specifically, through their own Python package.
Such a herculean task of accessing this data was not humanly possible. Retrieving the data is one thing, but making it readily available for labelling was another. Two noteworthy challenges they faced regarding the data set was the fact that the images used to be very big in size and that the images themselves had a different rotation—most of them were diagonal! But this did not stop the project team from pressing on forward.
Source: The Martian Chronicles — When Deep Learning meets Global Collaboration
Here’s a step-by-step look at how they did it, from Samir Sheriff , a member of the project team:
- Create HTTP Query String to fetch images from the ODE REST Interface
- Parse the Query Response XML and extract JP2 URLs
- Download all JP2 images using the URLs obtained in step 2
- Slice each JP2 image into smaller, equally-sized chunks
- Identify black margins in each chunk
- Remove black margins by rotating or cropping them out while retaining the original resolution of that chunk OR discarding the entire chunk if that is not possible
- (Optional) Aggregate all chunks and save them in a single NPZ file
Thankfully, you can perform all of these steps yourself by just installing one Jupyter Python package and running a couple of commands.
Close Encounters with the Dream TeamOnce the data was readily available, the project team set out on the task of labelling all the images. On average they labelled 300-400 images in about a week! From here they identified 7 types of anomalies: craters, dark dunes, slope streaks, bright dunes, impact ejecta, spiders and swiss cheese.
Source: The Martian Chronicles — When Deep Learning meets Global Collaboration
With the overwhelming amount of images to label, the project team’s size certainly proved that there is strength in numbers. But it’s not just the number of brains you’ve got, it’s also the skills and domain knowledge a diverse team brings to the table.
One of the project team members or collaborators enthused, “In a world being plagued by greed, hate, and intolerance, Omdena comes as a breath of fresh air to do away with national barriers. This project is a testament to the fact that bringing together a group of strangers from different corners of the Earth, who have never met each other before; transcending geographical borders and time zones to work together and solve fascinating social problems; whilst learning from and inspiring each other every single day, is not just a pipe dream.”
A World of PossibilitiesTeamwork truly makes the dream work and in their time together, collaborating from different locations worldwide, they were able to apply techniques like GANs (generative adversarial networks) to build an expert system that could classify images with an anomaly score. They were able to develop a Python package that made processing data from Mars not just possible, but also efficient. Finally, they were able to test different models and identify the best one: the U-Net model.
The Eskwelabs community can soon experience the magic of this Mars project through our collaboration with Omdena, where the original project team will work alongside our learners to reproduce the deep learning algorithms.
Erum Afzal, Omdena’s lead instructor in this collaboration, had words of encouragement for those who wish to embark on the same journey. She said, “We’ve already created pipelines. If you have images and you are worried about pre-processing, we already have the pipeline. Participants just have to run 3-4 lines of code and understand the data.”
She affirms that the algorithm they’ve developed doesn’t have to be just for Mars data. Erum says you can also take photos from the Moon, apply this pipeline in healthcare images (Ex: Brain cancer), and in agriculture (Ex: Which crops are anomalous? Which ones do we need to treat?).
“Just run the algorithm and access that data.”
Audio Credit: NASA/JPL-Caltech/DPA
What this Means for the PhilippinesOmdena hosted this AI challenge (aka their “Anomaly Detection on Mars Using Deep Learning” project) with the University of Bern in Switzerland back in 2019—one of the mentors on the challenge was from the NASA Frontier Development Lab (FDL). What the FSL does is they apply AI technology to space science in order to lead new research and innovate new tools which they hope will help solve challenges that humanity faces today.
Certainly, Omdena’s project hits close to home for Filipino data talent—whether these data scientists have a particular love for all things Astronomy or not—because there are a number of local applications for the algorithm developed in 2019.
If the data set used to be photos of the Martian surface, then in the Philippine context, our data set could possibly contain photos of agricultural land, forests under observation, and more.
Renzo Wee, an electronics engineer from Zamboanga told ABS-CBN back in 2019 that the satellite that they were building would allow the Philippine government to “survey agricultural crops, protected forests and other areas of concern.”
The Philippines, with its 7,000+ islands, has a vast swath of agricultural land and numerous fishing zones. It’s exciting to see how Omdena’s project can help improve our country’s ground-based monitoring, consequently impacting our agricultural production, the fishing yield of our fishermen, and the areas of forests we aim to preserve and protect.
Looking for E.T. with Filipino data talentOmdena have wrapped up their “Anomaly Detection on Mars Using Deep Learning” project back in 2019, but this is only just the beginning. We at Eskwelabs are excited to partner with Omdena in bringing this project to our Philippine audience via our newly-launched upskilling product, the Data Club.
Eskwelabs is an online data upskilling school in the Philippines with the mission of driving upward social mobility in the future of work through data skills education. Our past work from bootcamps and corporate training showed us project-based learning through cohorts produces transformative results for learners. We are creating the Data Club to democratize this experience as a digital place where people of varying data skill levels can come together to create hands-on data projects.
Data projects vary in terms of schedule, pre-requisites, instructors, mentors, and output. For this Eskwelabs x Omdena partnership, we will be turning the “Anomaly Detection on Mars Using Deep Learning” project into a data sprint titled “Looking for E.T.” and we are inviting those with intermediate Python backgrounds to join this dynamic learning experience.
Participants of this data sprint will leave the experience with portfolio-worthy output, connections with industry experts, and genuine relationships with fellow data enthusiasts. Data Club’s “Looking for E.T.” (aka Eskwelabs x Omdena’s “Anomaly Detection on Mars Using Deep Learning” project) sprint is now accepting interested participants.