Day 6/7 of training some proper realism into this plastic AI world.
Trained on 54 frames of open traffic webcams.
Most of them from NA, some from EU & Asia.
no captions
I think this LoRa is gonna be a very interesting one to learn from. Its quite bad at generating traffic lines, crosswalks, etc.. Also for some reason people come out a little funny walking.
Gonna spend some time to improve the dataset and training.
This one should have a proper hash from the get go, so post them generations and see some funny walks!
trigger word: trafficCam
strength: 1.8+