Model Card for davit_fl_base_florence-v2
A DaViT base image encoder from Microsoft's Florence-2 model, converted to the Birder format for image feature extraction. This version preserves the original vision backbone weights and architecture for downstream tasks.
See https://huggingface.co/microsoft/Florence-2-base for further details.
Model Details
-
Model Type: Image classification and detection backbone
-
Model Stats:
- Params (M): 90.4
- Input image size: 768 x 768
-
Papers:
- DaViT: Dual Attention Vision Transformers: https://arxiv.org/abs/2204.03645
- Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks: https://arxiv.org/abs/2311.06242
Model Usage
Image Embeddings
import birder
from birder.inference.classification import infer_image
# Option 1: manual setup (more control over preprocessing)
net, model_info = birder.load_pretrained_model("davit_fl_base_florence-v2", inference=True)
# Get the image size the model was trained on
size = birder.get_size_from_signature(model_info.signature)
# Create an inference transform
transform = birder.classification_transform(size, model_info.rgb_stats)
# Option 2: helper (quick start with default preprocessing)
net, model_info, transform = birder.load_pretrained_model_and_transform("davit_fl_base_florence-v2", inference=True)
image = "path/to/image.jpeg" # or a PIL image
out, embedding = infer_image(net, image, transform, return_embedding=True)
# embedding is a NumPy array with shape of (1, 1024)
Detection Feature Map
from PIL import Image
import birder
net, model_info, transform = birder.load_pretrained_model_and_transform("davit_fl_base_florence-v2", inference=True)
image = Image.open("path/to/image.jpeg")
features = net.detection_features(transform(image).unsqueeze(0))
# features is a dict (stage name -> torch.Tensor)
print([(k, v.size()) for k, v in features.items()])
# Output example:
# [('stage1', torch.Size([1, 128, 192, 192])),
# ('stage2', torch.Size([1, 256, 96, 96])),
# ('stage3', torch.Size([1, 512, 48, 48])),
# ('stage4', torch.Size([1, 1024, 24, 24]))]
Citation
@misc{ding2022davitdualattentionvision,
title={DaViT: Dual Attention Vision Transformers},
author={Mingyu Ding and Bin Xiao and Noel Codella and Ping Luo and Jingdong Wang and Lu Yuan},
year={2022},
eprint={2204.03645},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2204.03645}
}
@misc{xiao2023florence2advancingunifiedrepresentation,
title={Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks},
author={Bin Xiao and Haiping Wu and Weijian Xu and Xiyang Dai and Houdong Hu and Yumao Lu and Michael Zeng and Ce Liu and Lu Yuan},
year={2023},
eprint={2311.06242},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2311.06242},
}