Resultado da Busca
Há 18 horas · Yann LeCun summed it up in the second post: use convolution with stride or pooling at low level, and at high level Use a self-attention loop and use feature vectors to represent objects. He also bets that Tesla Fully Self-Driving (FSD) uses convolutions (or more complex local operators) at low levels and combines more at higher levels Global loop (possibly using self-attention).
Há 18 horas · 在Transformer大一统的时代，计算机视觉的CNN方向还有研究的必要吗？今年年初，OpenAI视频大模型Sora带火了VisionTransformer（ViT）架构。此后，关于ViT与传统卷积神经网络（CNN）谁更厉害的争论就没有断过。近日，一直在社交媒体上活跃的图灵奖得主、Meta首席科学家YannLeCun也加入了ViT与CNN之争的讨论。这 ...
Há 18 horas · 在Transformer大一統的時代，電腦視覺的CNN方向還有研究的必要嗎？今年年初，OpenAI視訊大模型Sora帶火了VisionTransformer（ViT）架構。此後，關於ViT與傳統卷積神經網路（CNN）誰更厲害的爭論就沒有斷過。近日，一直在社群媒體上活躍的圖靈獎得主、Meta首席科學家YannLeCun也加入了ViT與CNN之爭的討論。這件 ...
