Index
For more recent paper summaries, refer to Fully Connected at Weights & Biases. Here are some other blogs that I’ve written:
- Inside Hugging Face’s Accelerate!
- Explained: Characterizing Signal Propagation to Close the Performance Gap in Unnormalized ResNets
- Revisiting ResNets: Improved Training and Scaling Strategies
I am also involved in hosting fortnightly paper reading groups at W&B, refer here for more details.
Posts
-
The Annotated DETR
-
The sad state of AI and tech startups in Australia today and what can we do about it
-
“Adam” and friends
-
ViT - AN IMAGE IS WORTH 16X16 WORDS: TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE
-
The EfficientDet Architecture in PyTorch
-
EfficientDet - Scalable and Efficient Object Detection
-
U-Net: A PyTorch Implementation in 60 lines of Code
-
Top 100 solution - SIIM-ACR Pneumothorax Segmentation
-
GeM Pooling Explained with PyTorch Implementation and Introduction to Image Retrieval
-
SIIM-ISIC Melanoma Classification - my journey to a top 5% solution and first silver medal on Kaggle
-
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
-
Group Normalization
-
DenseNet Architecture Explained with PyTorch Implementation from TorchVision
-
Squeeze and Excitation Networks Explained with PyTorch Implementation
-
Label Smoothing Explained using Microsoft Excel
-
An introduction to PyTorch Lightning with comparisons to PyTorch
-
What is Focal Loss and when should you use it?
-
The Annotated GPT-2