https://t.co/ViQXaM2t8f
ICCV2021の、自動着色モデルのエッジ部分をきれいにするための手法が面白かったので紹介(画像は論文から引用)

正解画像と既存の事前学習モデルで着色した画像とでエッジを検出して、その差分を苦手境界として擬似的に学習に利用
推論時は人手で変になってる所をなぞる

4 9

Want to train deep networks to model 3D human pose and motion? AMASS is a new dataset with 45 hours of mocap - and growing – all in SMPL format. Tutorial code shows how to use AMASS for deep learning. To appear: data: https://t.co/FnHeQhM3C9

50 161

"If you gaze long into your food, the food also gazes into you"

One of my sweet memories from 2017 the amazing Italian food at the banquet in Arsenale.
These antipasti plates made a great good for and creations. Buon appetito 😉

2 11