SlideShare a Scribd company logo
Representation learning by learning to count
Representation learning by learning to count
Representation learning by learning to count
•
-
•
-
-
•
•
-
•
-
•
-
•
-
•
•
•
-
-
-
•
-
→気持ちを画像に掛けて、それを直接学習
→こういう気持ちが画像に掛かってるだろう、
という仮説に基づいて学習
•
•
•
-
-
-
-
-
Representation learning by learning to count
•
0.34
0.54
0.02
0.41
1.31
CNN
CNN
CNN
CNN
•
•
1.31
1.21
ここを⼀
致させる
•
-
•
-
-
•
•
•
•
•
-
-
•
•
•
-
•
-
-
Highest	counting Lowest	counting
•
-
•
•
•
•
•

More Related Content

PDF
A quantum computational approach to correspondence problems on point sets
PDF
F0-Consistent Many-to-many Non-parallel Voice Conversion via Conditional Auto...
PDF
YOLACT real-time instance segmentation
PDF
Product Managerの役割、周辺ロールとの差異
PDF
ChainerRLで株売買を結構頑張ってみた(後編)
PDF
Temporal Cycle Consistency Learning
PDF
PDF
20190414 Point Cloud Reconstruction Survey
A quantum computational approach to correspondence problems on point sets
F0-Consistent Many-to-many Non-parallel Voice Conversion via Conditional Auto...
YOLACT real-time instance segmentation
Product Managerの役割、周辺ロールとの差異
ChainerRLで株売買を結構頑張ってみた(後編)
Temporal Cycle Consistency Learning
20190414 Point Cloud Reconstruction Survey

More from Fujimoto Keisuke (20)

PDF
Chainer meetup 9
PDF
20180925 CV勉強会 SfM解説
PDF
Sliced Wasserstein Distance for Learning Gaussian Mixture Models
PDF
LiDAR-SLAM チュートリアル資料
PDF
Stock trading using ChainerRL
PDF
Cold-Start Reinforcement Learning with Softmax Policy Gradient
PDF
Dynamic Routing Between Capsules
PDF
Deep Learning Framework Comparison on CPU
PDF
ICCV2017一人読み会
PDF
Global optimality in neural network training
PDF
CVPR2017 oral survey
PDF
Point net
PDF
CVIM最先端ガイド6 幾何学的推定のための最適化手法 3.5 - 3.8
PDF
Value iteration networks
PDF
sublabel accurate convex relaxation of vectorial multilabel energies
PDF
Deep SimNets
PDF
Deep Fried Convnets
PDF
L0TV: a new method for image restoration in the presence of impulse noise
PDF
Learning to forget continual prediction with lstm
Chainer meetup 9
20180925 CV勉強会 SfM解説
Sliced Wasserstein Distance for Learning Gaussian Mixture Models
LiDAR-SLAM チュートリアル資料
Stock trading using ChainerRL
Cold-Start Reinforcement Learning with Softmax Policy Gradient
Dynamic Routing Between Capsules
Deep Learning Framework Comparison on CPU
ICCV2017一人読み会
Global optimality in neural network training
CVPR2017 oral survey
Point net
CVIM最先端ガイド6 幾何学的推定のための最適化手法 3.5 - 3.8
Value iteration networks
sublabel accurate convex relaxation of vectorial multilabel energies
Deep SimNets
Deep Fried Convnets
L0TV: a new method for image restoration in the presence of impulse noise
Learning to forget continual prediction with lstm
Ad