Hey, I’ve been looking at ways to auto‑select the best lighting angles for drone footage. I’d build a neural net that analyses real‑time RGB histograms and depth maps, then recommends the angle that maximizes contrast while keeping the subject’s exposure flat. The challenge is getting enough labeled data for “good” vs. “bad” shots, and making sure the model runs on the drone’s limited CPU. I think it could work, but I’ll probably keep tweaking the loss function until it’s just right.

Sounds like a solid plan, but remember the small dataset might bias the model too heavily toward those golden hours—don’t forget to inject some variability. Also, be careful with that warmth bias; if you over‑warm it, the color grading can look washed out. Keep iterating, tweak the loss, and you’ll get the drone’s light choreography working. Good luck!

Sounds good—just keep an eye on the learning rate; a small tweak can stop over‑fitting to that warm tint. Happy tinkering!

Nice! Just remember to log the sensor data each time; the more telemetry you keep, the faster the model learns. Keep hacking!

Glad to hear it, just make sure the telemetry includes timestamps and exposure values so the network can correlate lighting conditions. Keep iterating!