AL Zobaer

AL Zobaer

M2 researcher in scale-invariant 3D plant phenotyping, 3DGS, and autonomous agricultural robotics.

Welcome to my research blog

less than 1 minute read

Published:

This is the first post on my research blog. I plan to use this space to share short, informal notes on my work at the intersection of robotics, computer vision, and precision agriculture — the ideas behind the papers, things that did not make it into them, and lessons learned along the way.

Four stage pipeline: photograph the plant from many angles, rebuild it in 3D, measure traits, repeat daily to build a growth curve.
Most of what I work on is some part of this loop: photograph a plant, rebuild it in 3D, measure it, and do it again tomorrow — without ever cutting anything down.

Some topics I hope to write about:

  • Scale-invariant phenotyping — measuring plant traits in normalized render-space, where perspective projection cancels the per-session scale ambiguity of Structure-from-Motion.
  • 3D Gaussian Splatting for high-fidelity crop reconstruction, and its extension toward 4D (temporal) growth.
  • Autonomous agricultural robotics — low-cost, self-charging ground vehicles for operator-free data acquisition in greenhouses.

Thanks for reading. More soon.

You May Also Enjoy

COLMAP in practice

13 minute read

Published:

Two hundred tourists photograph a statue, and from the photographs alone you can work out where each of them stood. That is Structure-from-Motion, and COLMAP is the program that does it — explained from scratch, then in practice.

実務としての COLMAP

2 minute read

Published:

二百人の観光客が彫像を撮る。その写真だけから、誰がどこに立っていたかを言い当てられる。それが Structure-from-Motion であり、COLMAP はそれを行うプログラムです。ゼロから説明し、そのあと実務の話へ。

বাস্তবে COLMAP

12 minute read

Published:

দুইশো পর্যটক একটি ভাস্কর্যের ছবি তোলে, আর কেবল সেই ছবিগুলো দেখেই বলে দেওয়া যায় কে কোথায় দাঁড়িয়ে ছিল। এটাই Structure-from-Motion, আর COLMAP হলো সেই প্রোগ্রাম যা কাজটি করে — একেবারে গোড়া থেকে, তারপর হাতেকলমে।

What is a Gaussian Splat?

6 minute read

Published:

The usual way to turn a point cloud into a model is to stretch a surface over it. Plants punish that assumption — so here is the alternative.