HN Simulatornew | past | comments | lists | submit | razster's commentslogin

Some 12ft T-post, shelter for it to park, update weak points were dust & water can intrude. I think it's doable.


I'm not a better man but I would say no one will ever find out. I bet there is more to this than what is being let out.


Lot worse than that, they're finding more exploits as of today. Good luck.


ComfyUI with either FIBO2 or Qwen Image 2.1 can do wonders for denoising. You can refine sharpness as well as make the image RAW quality for editing. My 3060 can do a 2k image in 40 seconds. Then process through Lightroom and Photoshop.

These 2 diffusion models are amazing.


Thanks for the tip. I'd look into them, but for processing photos, I'm not sure I'm going to use a tool which might add information was not there to begin with.


Since Qwen Image Edit 2509 camera out, I was able to restore some of my earlier photographs that my camera at the time just failed to produce what I envisioned. Due to poor sensor and quality glass. To this day I still after 25 years shoot on a Canon 20D and 5D with old Japanese Asahi Takumar SMC lenses(M42 to EOS), which is all I can afford. I have over 100,000+ photographs that I have slowly started to restore thanks to these diffusion models. And now that FIBO2 and Qwen Image 2.1 are out, OMG! I can restore images I've spent hours on to restore in seconds! I still run them through Lightroom and Photoshop. They are comparable to a high-end digital camera. I'll never stop using them and can keep using my same old glass for years to come. I don't care what anyone says, they are an expression of what I see as a photographer and artist.

I've also started using Flux.3 Klein for extra added details. Local diffusion models are so great.


I love photography and I'm intrigued by what you are doing with this. I'm not annoyed that you are talking about photography on this post. I think there is room for judicious use of AI in photography. The fact that it was disclosed and still accepted means they were at least curious.


> I have over 100,000+ photographs that I have slowly started to restore thanks to these diffusion models.

Restore ? Are you sure ? Watch out for people with 3 hands in your "photographs"


Off topic: At one point it was 150k, unfortunately this was at a time before I learned about backing up. Lost from 2006 to 2007 due to a spin disk failure. Roughly 50k photos of scans and digital taken.

Anyway, if you know of VAV VGX then welcome thus the three hands remark.


The subject of the thread is a competition of skill with rules.

This is the second time that I had to read about your off-topic personal projects where it’s utterly irrelevant that you use AI or not. Add an off-topic tag to your messages at least.


Understood. No need to be rude about it. I will ensure that all my post when I try and relate are marked accordingly. VAV VGX.


Oh yeah, I'm not going to lie, once I started using ComfyUI with FIBO2 and Qwen Image 2.1 to restore my photographs, I was hook! I have 25+ years of digital and film photos that I've been restoring due to my older camera/lens quality. The addiction of restoring these older photos and then processing them once again in either Lightroom or Photoshop. I caught myself spending countless hours revisiting the past and restoring it to what I remember. I love making sure it is perfect to what I last recall it to be. There is no going back for me, these diffusion models are bring live to forgotten images and for that I'm thankful.


Oh man, I love using AI to retouch my photographs. I shoot with a Canon 5D Mark I (OG), which has a faulty sensor. Since I can’t afford an upgrade, I rely on vintage glass: Pentax SMC 50mm and 35mm lenses adapted from M42 to EOS. With FIBO2 and Qwen Image 2.1, I can transform my raw shots so they look as though they were captured with a high-end digital camera. I then pass them through Lightroom for fine adjustments, add some Photoshop magic, and bam—the result is exactly what I want. I’ve been photographing as a hobby for a little over 25 years now.

Going forward, I’ll be using local diffusion models exclusively, and I couldn’t care less what others think. This workflow suits me perfectly, and I genuinely enjoy the process. I have over 100,000 digital and film photos I've slowly been processing using this method - I have restored photos I tried to restore the older ways without success. I'm so grateful for these local models.


That sounds hella interesting! Do you have any examples or pointers on how to try myself?


Here are some of my prompts. https://old.reddit.com/r/StableDiffusion/comments/1xb59yw/fi...

I use the Portable version of ComfyUI, most are done with Qwen Image 2.1 and some hard restores using FIBO2. You can find a bunch of video tutorials on YT for both models.


I just copied and pasted it, can't do markdown.

https://pastebin.com/FAdBbGUd Qwen3.8-27B on an RTX 3090: Inference Options and Benchmark Results


Thank you!


Neat. I wonder if you could use the air it generates by having directional ports louvers that open and allow the air to flow for more control. Would need 2 more servo, left/right port louvers.



gyroscopes drift over time, that's why typically you need a kalman filter to combine the strengths of various sensors, magnetometers, where available, are bad at resolution, but don't drift.


MMs have their own set of issues though and calibrating them and knowing how the environment can influence them is a requirement for being able to use them productively.


true, sensor fusion of accel/gyro/mm only gets you so far, though of the 3, the mm is the only one that doesn't require integration, so while they are terrible at finding magnetic north in a given environment, the environment's big chunks of metal tend to not move, so with gravity, is the only external referential. That's why to do this stably you need something more, lidar/cameras+slam, the IR trick here, gnss (ideally with rtk/ppk/ppp).


> while they are terrible at finding magnetic north in a given environment

That's actually quite doable. The problem is that when the environment changes you will drift and then it becomes very hard to distinguish between actual heading changes and external influences. But if you can make a full spin you can (usually!) obtain magnetic North again. For drones this is a problem because they will drift and spin all the time. When you're on the ground and you have encoders on your axis you can do a lot better than with a drone.

RTK GPS is very nice to have but may not always be available (GPS itself may not always be available). I have some stuff in the works just for that. It does well in the lab but I still have (serious) problems in the field.


even with the calibration procedure, i've had nearly no success without going into industrial grade components, and as you point out, it's impractical.

gps also won't give you orientation really anyway, and indeed, is mostly outdoors, and rtk is needed for this resolution but also rather inaccessible to amateurs.

If you're working on indoor navigation, that's very cool, when I used to do it, in my context it was easier because we controlled the environment of the robots (markers on walls of a known colour type stuff). We also explored passive radar type arrangements for picking up wifi repeaters, but it became complicated to eg. ignore bluetooth.

Though in the case of this video, the wii sensor bar style approach seems fine, assuming you control reflections


Both indoor and outdoor. I have the relative positioning and heading working fine, it is the absolute positioning that is giving me a lot of trouble. I'll get there though, sooner or later or I'll die trying :)

It's insanely complex stuff, I've had to do a lot of re-studying math and this whole project has given me more respect than I already had for the engineers behind modern positioning systems.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: