Ivo Gasparini

TreeGPT: Generative Pre-trained Transformer for forestry applications with 3D point clouds

Sun, 10 Aug 2025 00:00:00 +0000

Figure 1: Example of a segmented point cloud of a forest in Canton Neuchâtel used for pre-training.

For my master’s thesis, I tackled a challenge that has captured my interest for years: leveraging LiDAR point clouds to generate automated forest inventories.

I remember when I was studying forest sciences that one of the recurring problems in forest planning was the lack of accurate data about the forest area in question. How to effectively plan the allowable cut, rotation period, tree species composition, and any silvicultural intervention without an accurate overview of what we are working with? Data in this field often comes from the national forest inventory, whose resolution is inadequate when considering individual forest areas. In the past, manual sample inventories were systematically organized by public entities on a widespread basis, but nowadays costs have become prohibitive for efforts of such magnitude.

The weird stuff going on in LLM-land

Wed, 17 Sep 2025 00:00:00 +0000

Scaling transformers (the leading model architecture in AI) to infinity and training them on basically the entirety of human written knowledge has led to the AI revolution with the LLMs we know and have mixed feelings about today. While what intelligence is and how it applies to LLMs is a topic far too vast for a single blog post, if we take LLMs at face value, they sure strike us with brilliance from time to time. What is really going on in these models? Is it just very high-dimensional pattern matching? Is the brilliance just a reflection of our own? Do we fundamentally function the same?

Pre-print: Automatic inventory of retaining walls from aerial lidar data using 3D deep learning

Sun, 14 Sep 2025 00:00:00 +0000

The paper has been accepted! Post coming soon…

Tinkering with numerical simulations: my attempt at an explicit solver for rockfall barrier analysis

Mon, 23 Jun 2025 00:00:00 +0000

Figure 1: Rockfall barrier in RBA during impact.

📄 RockBarrierAnalytica - Report

As part of my work at BFH, we inspect rockfall barriers sold in Switzerland to check their eligibility for federal subsidies and issue quality certifications. Rockfall barriers are mechanically complex systems due to the extreme dynamics of the processes involved and can behave unpredictably. To better understand the behavior of these systems, numerical simulations can be a powerful tool. However, existing commercial software solutions often come with limitations, such as high costs or lack of modeling flexibility.

How I built this site

Sun, 15 Jun 2025 00:00:00 +0000

I built this site to learn, share and connect. To share and connect, a blog format made sense. To learn, I wanted to build it from the ground up, or at least as much as reasonable. I also wanted to keep costs low, ideally below 5 CHF/month, which is about 1/3 of the price of most site builders but without any of the limitations.

Tech stack

For the frontend, I went with Hugo, a static site generator written in Go. I feel like static sites are underrated these days. Raw HTML from the server is fast and efficient. I picked the beautiful hugo-paper theme and modded it a bit (contact form, multi-language support, sticky header, and a bunch of other tweaks).

Contact

Wed, 01 Jan 2025 00:00:00 +0000

About me

Wed, 28 Aug 2019 00:00:00 +0000

I’ve always been obsessed with figuring out how things work. I believe that in life there is nothing to fear, only to understand. I believe technology should make life lighter, not the opposite.

I hold a bachelor in forest sciences and a master in data science. I have worked in government and in applied research. Check out my projects. In my free time I’m interested in music, science, the economy, and the occasional random topic that has triggered my obsession.

Tue, 28 May 2019 00:00:00 +0000