DOLG in torch and tensorflow (TF2)
Re-implementation (Non Official) of the paper DOLG: Single-Stage Image Retrieval with Deep Orthogonal Fusion of Local and Global Features accepted at ICCV 2021. paper
The pytorch checkpoint has been converted into tensorflow format (.h5) from this repository : https://github.com/feymanpriv/DOLG (Official)
Pipeline
Installation
pip install opencv-python==4.5.5.64
pip install huggingface-hub
to install dolg :
pip install dolg OR pip install -e .
Inference
To do some inference on single sample, you can use python script in examples/ folder or use as follows:
import dolg
import numpy as np
from dolg.utils.extraction import process_data
depth = 50
# for pytorch
import torch
from dolg.dolg_model_pt import DOLG
from dolg.resnet_pt import ResNet
backbone = ResNet(depth=depth, num_groups=1, width_per_group=64, bn_eps=1e-5,
bn_mom=0.1, trans_fun="bottleneck_transform")
model = DOLG(backbone, s4_dim=2048, s3_dim=1024, s2_dim=512, head_reduction_dim=512,
with_ma=False, num_classes=None, pretrained=f"r{depth}")
img = process_data("image.jpg", "", mode="pt").unsqueeze(0)
with torch.no_grad():
output = model(img)
print(output)
# for tensorflow
import tensorflow as tf
from dolg.dolg_model_tf2 import DOLG
from dolg.resnet_tf2 import ResNet
backbone = ResNet(depth=depth, num_groups=1, width_per_group=64, bn_eps=1e-5,
bn_mom=0.1, trans_fun="bottleneck_transform", name="globalmodel")
model = DOLG(backbone, s4_dim=2048, s3_dim=1024, s2_dim=512, head_reduction_dim=512,
with_ma=False, num_classes=None, pretrained=f"r{depth}")
img = process_data("image.jpg", "", mode="tf")
img = np.expand_dims(img, axis=0)
output = model.predict(img)
print(output)
Data
The model has been trained on google landmark v2. You can find the dataset on the official repository : https://github.com/cvdfoundation/google-landmark .
Citation :
@misc{yang2021dolg,
title={DOLG: Single-Stage Image Retrieval with Deep Orthogonal Fusion of Local and Global Features},
author={Min Yang and Dongliang He and Miao Fan and Baorong Shi and Xuetong Xue and Fu Li and Errui Ding and Jizhou Huang},
year={2021},
eprint={2108.02927},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
@misc{https://doi.org/10.48550/arxiv.2004.01804,
doi = {10.48550/ARXIV.2004.01804},
url = {https://arxiv.org/abs/2004.01804},
author = {Weyand, Tobias and Araujo, Andre and Cao, Bingyi and Sim, Jack},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {Google Landmarks Dataset v2 -- A Large-Scale Benchmark for Instance-Level Recognition and Retrieval},
Metadata
Release files for dolg 0.1.5.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| dolg-0.1.5.1.tar.gz | 15.1 kB | Without provenance | Details for dolg-0.1.5.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| dolg-0.1.5.1-py3-none-any.whl | Python 3 | none | any | Details for dolg-0.1.5.1-py3-none-any.whl |
Total release size: 33.7 kB
Release files / dolg-0.1.5.1.tar.gz
| Download URL | dolg-0.1.5.1.tar.gz |
|---|---|
| Size | 15.1 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
2cc4c6475479521c54853a55bedeb3da1437e6854def103a6fe58aca16471bd9
|
|
BLAKE2b-256 checksum How to use checksums |
2a807ee6e1fa67130ac85949e1eab15e9bc4ebeb720c8f3d504f2fad97a3161e
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.2 CPython/3.7.13
|
Release files / dolg-0.1.5.1-py3-none-any.whl
| Download URL | dolg-0.1.5.1-py3-none-any.whl |
|---|---|
| Size | 18.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
588e31d5a1f3e34db456ef8aa98a858e60ed3609d0b9d29966b2c10b79095b27
|
|
BLAKE2b-256 checksum How to use checksums |
816d3a3ed16afda47e2b3869dfbf4f1e7dfb9fc9975577d08a26528d38923de6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.2 CPython/3.7.13
|
Security for dolg 0.1.5.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.