AI search

Explore more awesome repositories

Describe what you need in plain English — the AI ranks thousands of curated open-source projects by relevance.

Convolutional Neural Networks - Ships a convolutional neural network backbone transferable to detection and segmentation tasks.

Depthwise Separable Convolutions - Uses 7x7 depthwise separable convolutions for efficient spatial context capture.

Hierarchical Stage Architectures - Organizes layers into four stages with increasing channel dimensions and decreasing spatial resolution.

Overlap Patch Embeddings - Replaces standard patch embedding with a convolutional layer processing overlapping image patches.

Pre-Activation Residual Blocks - Constructs residual blocks with activation and normalization applied before the convolution.

Learnable Scaling Layers - Implements learnable scaling factors on residual block outputs to stabilize deep network training.

Global Response Normalizers - Ships a global response normalization layer that normalizes feature maps across spatial dimensions.

Stochastic Depth Regularization - Applies stochastic depth regularization by randomly dropping residual blocks during training.

PyTorch Implementations - Implements the ConvNeXt architecture using the PyTorch framework for image recognition.

Strided Convolution Downsamplers - Implements strided 2x2 convolutions for spatial downsampling in the ConvNeXt architecture.

ImageNet Classifier Training - Provides training scripts for ConvNeXt models on ImageNet-1K and ImageNet-22K datasets.

Pre-trained Model Checkpoints - Provides pre-trained checkpoints and evaluation scripts for measuring top-1 and top-5 accuracy.

Model Fine-Tuning - Supports adapting pre-trained ConvNeXt models to custom datasets with training and logging.

Task-Specific Fine-Tuning Pipelines - Supports transferring the pre-trained ConvNeXt backbone to detection, segmentation, or custom tasks.

Pre-trained Model Zoos - Offers pre-trained model checkpoints ready for fine-tuning or direct inference on custom datasets.

Training and Evaluation Pipelines - Ships a training pipeline for ConvNeXt models on ImageNet-1K and ImageNet-22K datasets.

Vision - Ships scripts to adapt pre-trained ConvNeXt models to custom datasets with logging support.

Model Evaluation Benchmarks - Includes scripts to evaluate pre-trained ConvNeXt models on ImageNet validation data.

ConvNeXt