CRCV REU 2019 Kara Schatz
Background Math and Computer Science double major
Background Math and Computer Science double major Languages: Python Java C++ Scheme
Background Math and Computer Science double major Languages: Python Java C++ Scheme Computer Vision: General idea of Neural Networks
What I’ve Learned
What I’ve Learned Theoretical Convolution Boosting Neural Networks Activation Functions Gradient Descent Famous CNN Architectures Optical Flow Deep Learning
What I’ve Learned Theoretical Implementation Convolution Boosting Neural Networks Activation Functions Gradient Descent Famous CNN Architectures Optical Flow Deep Learning Keras VGG ResNet Pytorch C3D I3D
Classification on cifar100 Networks Used 3 layer CNN 4 layer CNN VGG ResNet
Classification on cifar100 Best: VGG 24.1% Validation Accuracy 51.2% Top 5 Validation Accuracy 3.304 Loss Using training dataset with 75 samples per class
Classification on cifar100 Best: VGG 24.1% Validation Accuracy 51.2% Top 5 Validation Accuracy 3.304 Loss Using training dataset with 75 samples per class
ResNet with varying Dataset Sizes
Project: Self-Supervized Cross-View Action Synthesis Dr. Yogesh Rawat
Project: Self-Supervized Cross-View Action Synthesis Datasets:
Project: Self-Supervized Cross-View Action Synthesis Datasets: NTU-RGBD 60K videos 3 views
Project: Self-Supervized Cross-View Action Synthesis Datasets: NTU-RGBD 60K videos 3 views Datasets: Panoptic dataset 65 video sequences ~540 viewpoints
Project: Self-Supervized Cross-View Action Synthesis