Guille (@GSBrread) header
Guille (@GSBrread) profile picture

computer vision for defense stuff | ars longa, vita brevis | art and some pics as well

Joined March 2015
Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)Photo by Guille (@GSBrread)
Saw @marshallrichrds drone adventures (ty for the inspiration) and had to give it a try myself. Reversed the E58's video feed, vision-based navigation next 🚁
3
3
21
933
Wrote about torch.compile and its predecessors last week, you can check it out here (or not) guillesanbri.com/pytorch-com…
58
🇬🇧 We are looking for people to join the team, if facing hard problems sounds like fun, reach out! Imho it's a great opportunity to work on computer vision in real-world, high-stakes systems. 🇪🇸 Dale vente hazme caso.
Hey everyone, We are growing at Indra Defense AI and we are looking for people with ML and CV backgrounds, from PhD level to early-career engineers. Desde Indra Defensa AI estamos buscando perfiles en ML/CV con diferentes nivel de experiencia, desde PhD hasta early-career.
1
3
133
I'm observing a mini Moravec's paradox within robotics: gymnastics that are difficult for humans are much easier for robots than "unsexy" tasks like cooking, cleaning, and assembling. It leads to a cognitive dissonance for people outside the field, "so, robots can parkour & breakdance, but why can't they take care of my dog?" Trust me, I got asked by my parents about this more than you think ... The "Robot Moravec's paradox" also creates the illusion that physical AI capabilities are way more advanced than they truly are. I'm not singling out Unitree, as it applies widely to all recent acrobatic demos in the industry. Here's a simple test: if you set up a wall in front of the side-flipping robot, it will slam into it at full force and make a spectacle. Because it's just overfitting that single reference motion, without any awareness of the surroundings. Here's why the paradox exists: it's much easier to train a "blind gymnast" than a robot that sees and manipulates. The former can be solved entirely in simulation and transferred zero-shot to the real world, while the latter demands extremely realistic rendering, contact physics, and messy real-world object dynamics - none of which can be simulated well. Imagine you can train LLMs not from the internet, but from a purely hand-crafted text console game. Roboticists got lucky. We happen to live in a world where accelerated physics engines are so good that we can get away with impressive acrobatics using literally zero real data. But we haven't yet discovered the same cheat code for general dexterity. Till then, we'll still get questioned by our confused parents.
143
525
109
2,468
399,589
Here are the first images from our forest mission 🌳🌴🌲, Biomass — launched less than two months ago: esa.int/Applications/Observi… (🧵 More images in the replies) 📸 Bolivia This image was created using Biomass's radar instrument and its different polarisation channels, with each colour revealing distinct characteristics of the landscape: Green = mainly rainforest Red = forested floodplains and wetlands Blue–purple = grasslands Black = rivers and lakes
5
68
19
226
207,050
Literally me (748274 hours left at the airport) More pics: guillesanbri.com/palettes/at…
2
200
atleast 5 more of the best selling "toy" drone models are now supported in turbodrone. I reverse engineered another mobile app called WiFi UAV that they all use and added it to the API. go check it out now!
22
42
4
648
33,660
It can all be so much more beautiful. Even the simple things... the soaps, the sinks, the lamps, and even the door knobs. All of it could delight us, if only we cared a little more. The problem isn't a lack of money. The problem is apathy. Making something beautiful begins with the decision to elevate your standards and commit to making something beautiful, and our culture has forgotten that simple fact. Until we raise our sights again, we'll be trapped in a dull existence of flat surfaces, lifeless white walls, and mass-produced objects that have no soul.
108
353
61
3,704
275,173
Bella Italia
1
4
179
Published a quick write-up on some experiments I did with DyT (Dynamic Tanh) replacing LayerNorm in small Vision Transformers a couple of weeks ago. Main insights ⬇️
2
1
178
LayerNorm was faster with small dimensions! DyT becomes faster than both LayerNorm and RMSNorm at large scales (e.g., Llama-sized inputs).
1
109
Now playing: Oops I did it again
Absolutamente lockeado in (correr por las mañanas mi pasión) (no)
1
130
Absolutamente lockeado in (correr por las mañanas mi pasión) (no)
1
2
228
Built a color palette generator using k-means to cluster pixels. The plan is to do this for every trip, but the real question remains: can I be arsed to do it for past pictures (not likely)
1
1
113
Anyway the thumbnails are messing up the aesthetics here are some pictures from Rome: guillesanbri.com/palettes/ro…
75
Morning routine (5:34 AM) -> Just published a post going over some basic guidelines for CUDA kernel benchmarking, the reasoning behind them, and some data. Check it out! Link in the response 👇🔽⬇️📉🕳️
1
2
145