README.md

metadata

datasets:
  - gaunernst/ms1mv3-recordio
library_name: timm
tags:
  - image-feature-extraction
  - timm

Model card for gaunernst/vit_tiny_patch8_112.cosface_ms1mv3

A Vision Transformer (ViT) for face recognition, trained on MS1MV3 dataset. The model was trained using this repo: https://github.com/gau-nernst/timm-face. It is fully compatible with timm.

Usage

import timm
import torch.nn.functional as F

model = timm.create_model("hf_hub:gaunernst/vit_tiny_patch8_112.cosface_ms1mv3", pretrained=True).eval()
embs = model(torch.randn(1, 3, 112, 112))  # output shape (1, 512)
embs = F.normalize(embs, dim=1)  # model output is not normalized