Lucrece (Jahyun) Shin·Jan 28, 2022Ch 10: Vision Transformer Part II — Iterative Erasing of Unattended Image Regions in PyTorchHelping the model better detect objects in images by iteratively erasing (i.e. darkening) regions of the image unattended by ViT using its…A response icon1A response icon1
Lucrece (Jahyun) Shin·Jan 28, 2022Ch 9. Vision Transformer Part I— Introduction and Fine-Tuning in PyTorchHow using self-attention for image classification reduces inductive bias inherent to CNNs including translation equivariance and locality…A response icon2A response icon2
Lucrece (Jahyun) Shin·Dec 23, 2021Ch 8. Adversarial Discriminative Domain Adaptation (ADDA): Quest for Semantic AlignmentOptimizing domain adaptation through toggling data annotation, training frameworks, and pre-training datasets
InCodeXbyLucrece (Jahyun) Shin·Oct 19, 2021Ch 7. Decoding Black Box of CNNs using Feature Map VisualizationsHow to ask CNN architectures useful questions to get insights about their behaviours
Lucrece (Jahyun) Shin·Oct 10, 2021Ch 6. Optimizing Data for Flexible Image RecognitionHow can we adjust input data and labels to encourage neural networks to “perceive” images flexibly as humans do?A response icon1A response icon1
Lucrece (Jahyun) Shin·Sep 22, 2021Ch 5. t-SNE Plots as a Human-AI Translatort-SNE Plots as a means of communicating with a deep learning model
Lucrece (Jahyun) Shin·Sep 21, 2021Ch 4. Transfer Learning with ResNet50 Part II- Model Analysis to Unexpected RiddleThinking about the Procedure >> Following the ProcedureA response icon1A response icon1
Lucrece (Jahyun) Shin·Sep 18, 2021Ch 3. Transfer Learning with ResNet50 Part I — from Dataloaders to TrainingSeed of Thought : Just how much about the ML model do we know after looking at the confusion matrix?A response icon2A response icon2
Lucrece (Jahyun) Shin·Sep 15, 2021Ch 2. Iterative Data CollectionHow to creatively design data for your ML problemA response icon1A response icon1
InCodeXbyLucrece (Jahyun) Shin·Sep 12, 2021DATA, DATA, DATA.Machine Learning is NOT magic.A response icon1A response icon1