Recent content by Airwolf

I haven't been able to produce what I'd call a good Nadir map. I get areas that don't line up quite right or are "ghosted" twice, areas that look like snow globes and I have vertical bands where the images appear darker. It is a fairly heavily treed and brushy plot with open areas and trails...

I picked up an AL1 spotlight, pretty nice! Some demo footage ...

Oh wow I looked closer and I think you're right! I have poor eyesight, thought it was just the screen surface when I first looked :cool:

Yeah, I have a ton of RAM, 256GB. The most I've seen WebODM sucking down so far is around 220GB ... it's a greedy beast! Been working through some errors I'm getting with this dataset, trying just an ortho run now. I'd like to try out DJI Terra but I don't have a Nvidia GPU. I think I saw an...

I did my first big ortho capture with the M4E this weekend. This thing is great, did ~100 acres @ GSD 2 with one battery. ~3,200 pics over 35 mins, returned @ 16%. Winds about 7-12mph. I started running it through WebODM with: If this even completes I'm expecting at least several days ...

Anyone have any recommendations for RC screen protectors? I got this from Amazon. It's ok but it's a fingerprint nightmare and it seems to dim the screen noticeably.

Another thought I had .. if I can get this process nailed down I might look at making a web front end other people can use. That way you could just upload images and get the necessary files back, no special knowledge needed.

It seems little bit like a beta right now, so hopefully the future holds more

White-tailed deer. At the moment it’s a manual process for DJI to load models. You have to send them the processed data, config file used and the pictures used for training. They presumably do some validation, possibly further processing, and then make it available for your drone.

I would say it is probably going to be a long road if you’re not familiar with this kind of stuff. If I get this working maybe I can write a guide.

Just the creator. That brings up an interesting point though. It would be neat if you could choose to share your models with select accounts, or the public. I can see practical applications like SAR.

So the gist of it is you have to do the training locally with a patched version of GitHub - open-mmlab/mmyolo: OpenMMLab YOLO series toolbox and benchmark. Implemented RTMDet, RTMDet-Rotated,YOLOv5, YOLOv6, YOLOv7, YOLOv8,YOLOX, PPYOLOE, etc. using the YOLOv8 model. Then you send the data to...

I got access. Will need to digest this, but there is more here than just supplying images.

I submitted a form to get access to algorithm development. Will let you know what I find out.

Oh, I like the concept but mine is a 15.7".