shipped
Fashion Recommender
ResNet18 visual similarity search for fashion items.
PythonPyTorchResNet18
Fashion Recommender is a content-based visual similarity search over a catalogue of fashion images. It encodes each item with a ResNet18 backbone and retrieves the nearest neighbours in embedding space, reaching a Top-5 accuracy of 0.942.
Image embeddings are pre-computed and cached, so retrieval at query time is a fast nearest-neighbour lookup rather than a full forward pass over the catalogue.