README.md

metadata

tags:
  - image-classification
  - timm
library_name: timm
license: apache-2.0
datasets:
  - imagenet-1k
  - unknown-6m

Model card for nextvit_small.bd_ssld_6m_in1k

A Next-ViT image classification model. Trained by paper authors on an unknown 6M sample dataset and ImageNet-1k using SSLD distillation.

Model Details

• Model Type: Image classification / feature backbone
• Model Stats:
  • Params (M): 31.8
  • GMACs: 5.8
  • Activations (M): 17.6
  • Image size: 224 x 224
• Pretrain Dataset: Unknown-6M
• Dataset: ImageNet-1k
• Papers:
  • Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarios: https://arxiv.org/abs/2207.05501
• Original: https://github.com/bytedance/Next-ViT

Model Usage

Image Classification

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model('nextvit_small.bd_ssld_6m_in1k', pretrained=True)
model = model.eval()

# get model specific transforms (normalization, resize)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # unsqueeze single image into batch of 1

top5_probabilities, top5_class_indices = torch.topk(output.softmax(dim=1) * 100, k=5)

Feature Map Extraction

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model(
    'nextvit_small.bd_ssld_6m_in1k',
    pretrained=True,
    features_only=True,
)
model = model.eval()

# get model specific transforms (normalization, resize)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # unsqueeze single image into batch of 1

for o in output:
    # print shape of each feature map in output
    # e.g.:
    #  torch.Size([1, 96, 56, 56])
    #  torch.Size([1, 256, 28, 28])
    #  torch.Size([1, 512, 14, 14])
    #  torch.Size([1, 1024, 7, 7])

    print(o.shape)

Image Embeddings

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model(
    'nextvit_small.bd_ssld_6m_in1k',
    pretrained=True,
    num_classes=0,  # remove classifier nn.Linear
)
model = model.eval()

# get model specific transforms (normalization, resize)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # output is (batch_size, num_features) shaped tensor

# or equivalently (without needing to set num_classes=0)

output = model.forward_features(transforms(img).unsqueeze(0))
# output is unpooled, a (1, 1024, 7, 7) shaped tensor

output = model.forward_head(output, pre_logits=True)
# output is a (1, num_features) shaped tensor

Model Comparison

By Top-1

model	top1	top1_err	top5	top5_err	param_count
nextvit_large.bd_ssld_6m_in1k_384	86.542	13.458	98.142	1.858	57.87
nextvit_base.bd_ssld_6m_in1k_384	86.352	13.648	98.04	1.96	44.82
nextvit_small.bd_ssld_6m_in1k_384	85.964	14.036	97.908	2.092	31.76
nextvit_large.bd_ssld_6m_in1k	85.48	14.52	97.696	2.304	57.87
nextvit_base.bd_ssld_6m_in1k	85.186	14.814	97.59	2.41	44.82
nextvit_large.bd_in1k_384	84.924	15.076	97.294	2.706	57.87
nextvit_small.bd_ssld_6m_in1k	84.862	15.138	97.382	2.618	31.76
nextvit_base.bd_in1k_384	84.706	15.294	97.224	2.776	44.82
nextvit_small.bd_in1k_384	84.022	15.978	96.99	3.01	31.76
nextvit_large.bd_in1k	83.626	16.374	96.694	3.306	57.87
nextvit_base.bd_in1k	83.472	16.528	96.656	3.344	44.82
nextvit_small.bd_in1k	82.61	17.39	96.226	3.774	31.76

Citation

@article{li2022next,
  title={Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarios},
  author={Li, Jiashi and Xia, Xin and Li, Wei and Li, Huixia and Wang, Xing and Xiao, Xuefeng and Wang, Rui and Zheng, Min and Pan, Xin},
  journal={arXiv preprint arXiv:2207.05501},
  year={2022}
}