LibreVGG16-cls

VGG-16 image classifier (224px, ImageNet-1k, 1,000 classes), repackaged for LibreYOLO. 138.4M parameters.

Source

Derived from pytorch/vision at commit 10f68dbd78b9aa5cab9328f3b2e99cfb0b608122, weight enum VGG16_Weights.IMAGENET1K_V1, official file vgg16-397923af.pth. Copyright (c) Soumith Chintala 2016. The releasing project's code is licensed under the BSD 3-Clause License.

The checkpoint object has no separate publisher-provided license file. This repository applies BSD-3-Clause on the explicitly disclosed basis implied by the releasing project, not as a confirmed checkpoint-specific grant. Torchvision warns that pretrained-model terms can depend on training data and users must determine permission for their use case. ImageNet is a separate dataset and is not distributed here.

Modifications

Learned parameters are unchanged. The checkpoint is metadata-wrapped into the LibreYOLO format (model_family, task, nc, names, and schema fields). LibreYOLO's native VGG graph retains the official module layout and produces bit-exact logits (max_abs_diff == 0). See weights/convert_vgg_weights.py in the LibreYOLO source repository.

Usage

from libreyolo import LibreYOLO

model = LibreYOLO("LibreVGG16-cls.pt")
result = model.predict("image.jpg")[0]
print(result.probs.top1, result.probs.top5)

License

BSD 3-Clause License, applied to this checkpoint on the disclosed implied basis above. See LICENSE and NOTICE.