Model card for levit 256.fb dist in1k A LeViT image classification model using convolutional mode (using nn.Conv2d and nn.BatchNorm2d). Pretrained on ImageNet 1k using distillation by paper authors. Model Details Model Type: Image classification / feature backbone Model Stats: Params (M): 18.9 GMACs: 1.1 Activations (M): 4.2 Image size: 224 x 224 Papers: LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference: https://arxiv.org/abs/2104.01136 Original: https://github.com/facebookresearch/LeViT Dataset: ImageNet 1k Model Usage Image Classification Image Embeddings Model Comparison model top1 top5 param count img size levit 384.fb dist in1k 82.596 96.012 39.13 224 levit conv 384.fb dist in1k 82.596 96.012 39.13 224 levit 256.fb dist in1k 81.512 95.48 18.89 224 levit conv 256.fb dist in1k 81.512 95.48 18.89 224 levit conv 192.fb dist in1k 79.86 94.792 10.95 224 levit 192.fb dist in1k 79.858 94.792 10.95 224 levit 128.fb dist in1k 78.474 94.014 9.21 224 levit conv 128.fb dist in1k 78.474 94.02 9.21 224 levit 128s.fb dist in1k 76.534 92.864 7.78 224 levit conv 128s.fb dist in1k 76.532 92.864 7.78 224 Citation
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy