A Floristic Survey of Viharamahadevi Park
A Preliminary Report on a Work in Progress
A Walk in the Park
I like to start my day with a bit of exercise, and these days this takes the form of walking about Viharamahadevi Park, Colombo.
As you might expect, on these morning walks, I see many interesting and beautiful examples of trees, shrubs, flowers and even weeds. Some, like Mango, Jac, or Jasmine, I’d recognize instantly. Others, sadly, were more alien.
I wondered if there was an easy way for a layperson (like me) to identify the plants, learn is scientific and common names.
After some research, I found PlantNet, which is a citizen science project, which uses Artificial Intelligence, Computer Vision, and Machine Learning, to identify plants from photos of them. And thanks to PlantNet, I was instantly able to identify most of the plants I didn’t know. All I needed to do was to take a photo of the plant with my phone, and then upload it to PlantNet.
[If you are interested in learning more about PlantNet, you can read my article, Identifying Plants with AI]
From One Plant to Many Plants
But then I wondered if I could go a step further. Why just one plant? What if I take pictures of all the plants at the park? It would make an interesting analysis, wouldn’t it?
I wondered if I could create a map of Viharamahadevi Park, which showed where every plant was, along with other information, like its species. The additional piece of information that a map needed was the location of the plant. How was I to get this?
One option was to use some location App on my phone to get the Latitude and Longitude of every plant I snapped a photo of. But then I realized there was a much easier method.
Every photo that my phone took, was automatically tagged with the GPS Coordinates of the location of the camera, and other information like the direction of the camera, and its focal length. This information is defined in the Exchangeable image file format Standard, or EXIF.
Thanks to PlantNet, I knew what the plant was. And now, thanks to EXIF, I also knew, where the plant was.
The Data
In the last couple of weeks, I’ve taken about 1,500 photos of Viharamahadevi Park’s Plants. Of these, PlantNet has been able to confidently identify about 1,000 plants. This has enabled me to build the “Map” I was talking about.
If you want to explore the map, you can use my Plants App. Please access the app on Desktop, as it uses a lot of data.
Some Analysis
Based on this data, the 10 most common Plant Species at Viharamahadevi Park are:
- Mesua ferrea (Appropriately, Sri Lanka’s National Tree. Also known as, Ceylon Ironwood or Naa.)
- Terminalia arjuna (Kumbuk)
- Mangifera indica (Amba, or Mango)
- Terminalia catappa (Kottamba, or Indian Almonds)
- Pongamia pinnata (Karanda, or Indian Beech)
- Artocarpus heterophyllus (Kos, or Jac)
- Cassia fistula (Ehela, or Indian Laburnum)
- Tectona grandis (Teak)
- Peltophorum pterocarpum (Maara)
- Tecoma stans (Kelanitissa, or Yellow Trumpet Flower)
Interestingly, my “Top 10” closely matched the results of an academic study by Madurapperuma et al, from 2015, which shows that my somewhat adhoc and lay-person approach, is not too far, in terms of results, from a more rigourous and scholastic one.
The data has also enabled me to create some interesting visualizations of the plants at the park. For example, here is a Sunburst Diagram of the taxonomic hierarchy of plants, weighted by frequency. This diagram, which is effectively a “Family Tree” of plants, helps you see how plants are related to others.
Next Steps
My immediate next step is to complete my survey of Viharamahadevi Park, as there are still a few parts of the park, which I haven’t covered.
Next, I plan to use my data to improve the quality of plant identification by the various AI models. Identification accuracy is still poor, for some species of plant.
After that, I might extend this exercise to other locations.
Let me know what you think. I’d be particularly interested to hear your thoughts on what I should do next.
If you’d like to see examples of identified plants, you could follow the Twitter/X account, @lk_plants (which is an automated bot, which posts photos of plants, along with their identification).
(For Data Scientists, and Plant Scientists)
You can find the dataset that I’ve collected thus far, at https://github.com/nuuuwan/lk_plants/tree/main/data.
(For Programmers)
You can find the code which I used for all of the above, at https://github.com/nuuuwan/lk_plants.
The code for the Plants App is at https://github.com/nuuuwan/plants.
