dino
约 1 分钟阅读
标签：无监督学习
- facebookresearch/dinov2GitHubfacebookresearch/dinov2facebookresearch/dinov2 · Apache-2.0, PyTorch · 2023-04-14 · Depth Estimation · Semantic Segmentation · Instance Retrieval笔记：dino
- Apache-2.0, PyTorch
- 2023-04-14
- Depth Estimation
- Semantic Segmentation
- Instance Retrieval
- Dense Matching - Consistently map all parts of an image without supervision.
- Sparse Matching - Compare DINOv2 patch features across two images to match their most similar parts.
- DINOv2: Learning Robust Visual Features without Supervision
- 用于特征提取，无监督
- 参考
- dinov2.metademolab.com/demos
- HF:docs/transformersHuggingFaceHF:docs/transformershuggingface.co · huggingface.co/docs/transformers/en/model_doc/dinov2
import torch
import dinov2.eval.segmentation.models
dinov2_vits14_reg = torch.hub.load('facebookresearch/dinov2', 'dinov2_vits14_reg')关联信息
反向链接、本文链接的其他页面和外部资料。
反向链接
- DINO笔记 · DINO v2
Self-Supervised Vision Transformers with DINO · 2021-04 · https://arxiv.org/abs/2104.14294 · facebookresearch/dino · DINOv2: Learning Robust Visual Features without Supervision
参考资料
GitHub1 条
- facebookresearch/dinov2github.com/facebookresearch/dinov2
HuggingFace1 条
- HF:docs/transformershuggingface.co/docs/transformers/en/model_doc/dinov2
其他外链1 条
- dinov2.metademolab.com/demosdinov2.metademolab.com/demos