lxztju/pytorch_classification
A complete PyTorch image classification toolkit supporting CNN architectures (ResNet, DenseNet, EfficientNet, Swin Transformer), knowledge distillation, model ensemble, and deployment via Flask API or C++ libtorch.
Velocity · 7d
+0.7
★ / day
Trend
→steady
star history
The repository provides end-to-end image classification capabilities using PyTorch and torchvision models. It includes training with warmup cosine learning rate scheduling, label smoothing, and multi-model ensemble prediction. The project supports CNN feature extraction followed by classification using SVM, Random Forest, MLP, or KNN classifiers. For deployment, it offers Flask + Redis cloud API, C++ libtorch inference, and TensorRT C++ inference options.