Elon Musk

@samiralyateem @WholeMarsBlog All frames stitched to single frame (tricky, as all cameras overlap in different ways) & then creating video segments for labeling & training. Everything from the labeling tools themselves to training to inference had to be rewritten.