Model card for regnetz_d8_evos.ch_in1k

A RegNetZ image classification model. Trained on ImageNet-1k by Ross Wightman in timm.

These RegNetZ B / C / D models explore different group size and layer configurations and did not follow any paper descriptions. Like EfficientNets, this architecture uses linear (non activated) block outputs and an inverted-bottleneck (mid block expansion).

• B16 : ~1.5GF @ 256x256 with a group-width of 16. Single layer stem.
• C16 : ~2.5GF @ 256x256 with a group-width of 16. Single layer stem.
• D32 : ~6GF @ 256x256 with a group-width of 32. Tiered 3-layer stem, no pooling.
• D8 : ~4GF @ 256x256 with a group-width of 8. Tiered 3-layer stem, no pooling.
• E8 : ~10GF @ 256x256 with a group-width of 8. Tiered 3-layer stem, no pooling.

This model uses customized EvoNorm-S0 normalization-activation layers instead of BatchNorm with SiLU activations.

This model architecture is implemented using timm's flexible BYOBNet (Bring-Your-Own-Blocks Network).

BYOBNet allows configuration of:

• block / stage layout
• stem layout
• output stride (dilation)
• activation and norm layers
• channel and spatial / self-attention layers

...and also includes timm features common to many other architectures, including:

• stochastic depth
• gradient checkpointing
• layer-wise LR decay
• per-stage feature extraction

Model Details

• Model Type: Image classification / feature backbone
• Model Stats:
  • Params (M): 23.5
  • GMACs: 4.5
  • Activations (M): 24.9
  • Image size: train = 256 x 256, test = 320 x 320
• Papers:
  • Fast and Accurate Model Scaling: https://arxiv.org/abs/2103.06877
  • Evolving Normalization-Activation Layers: https://arxiv.org/abs/2004.02967
• Dataset: ImageNet-1k
• Original: https://github.com/huggingface/pytorch-image-models

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('regnetz_d8_evos.ch_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(
    'regnetz_d8_evos.ch_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, 64, 128, 128])
    #  torch.Size([1, 64, 64, 64])
    #  torch.Size([1, 128, 32, 32])
    #  torch.Size([1, 256, 16, 16])
    #  torch.Size([1, 1792, 8, 8])

    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(
    'regnetz_d8_evos.ch_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, 1792, 8, 8) shaped tensor

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

Model Comparison

Explore the dataset and runtime metrics of this model in timm model results.

For the comparison summary below, the ra_in1k, ra3_in1k, ch_in1k, sw_*, and lion_* tagged weights are trained in timm.

Citation

@misc{rw2019timm,
  author = {Ross Wightman},
  title = {PyTorch Image Models},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  doi = {10.5281/zenodo.4414861},
  howpublished = {\url{https://github.com/huggingface/pytorch-image-models}}
}

@InProceedings{Dollar2021,
  title = {Fast and Accurate Model Scaling},
  author = {Piotr Doll{'a}r and Mannat Singh and Ross Girshick},
  booktitle = {CVPR},
  year = {2021}
}

@article{liu2020evolving,
  title={Evolving normalization-activation layers},
  author={Liu, Hanxiao and Brock, Andy and Simonyan, Karen and Le, Quoc},
  journal={Advances in Neural Information Processing Systems},
  volume={33},
  pages={13539--13550},
  year={2020}
}