Downloads · 30 days
0
sinder
haoqiwang/sinder
sinder is a feature extraction model from haoqiwang. Use it when you need embeddings to search or compare text. The card lists the license as apache-2.0.
Access
Public
Updated Oct 7, 2024
Repo size
4.5 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pth4.5 GB · 100%
At a glance
- Task
- Feature Extraction
- License
- apache-2.0
- Access
- Public
- Created
- Sep 22, 2024
- Updated
- Oct 7, 2024
- SHA
- 093acaa3
Model card
From the Hugging Face model README
Official code for SINDER: Repairing the Singular Defects of DINOv2 (ECCV 2024 Oral)
Citation
@InProceedings{Haoqi_2024_ECCV,
author = {Wang, Haoqi and Zhang, Tong and Salzmann, Mathieu},
title = {{SINDER}: Repairing the Singular Defects of DINOv2},
booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},
month = {September},
year = {2024}
}
Install
conda env create -f environment.yml
conda activate sinder
pip install -e .
Train
Put ImageNet-1K dataset at data/imagenet folder.
./main.py
Visualize
# visualize original dinov2
./visualize.py resources/example.jpg
# visualize sinder, ckpt download link below
./visualize.py resources/example.jpg --checkpoint path/to/sinder.pth
Checkpoints
- Task
- Feature Extraction
- License
- apache-2.0
- Created
- Sep 22, 2024
- Updated
- Oct 7, 2024