turhancan97/visbench-classification-dinov2_vits14

classification probe for dinov2_vits14

A trained probe head, not a backbone. It is the small module VisBench fits on top of frozen dinov2_vits14 features to measure what those features carry.

import visbench
from visbench.hub import load_probe_from_hub

backbone = visbench.get_backbone("dinov2_vits14")
probe = load_probe_from_hub("turhancan97/visbench-classification-dinov2_vits14", backbone=backbone)

It only works with this backbone

These weights were fitted on features from dinov2_vits14, taken with pooling=cls and feature_mode=dense_only. Loading them against anything else is refused, because the failure is otherwise silent: a head fitted on one pooling and fed another has the right shapes and produces a plausible, wrong number.

backbone	dinov2_vits14
backbone key	dinov2/dinov2_vits14/224/7764ea0f912e
task	classification (high_level)
pooling	cls (requested default)
feature mode	dense_only
layers	None

Reported scores

metric	value
top1	0.9939
top5	0.9997

Reproducing it

Fitted with:

• batch_size: 256
• epochs: 200
• lr: 0.01
• num_classes: 10
• optimizer: adamw
• standardize: False
• weight_decay: 0.0001

Generated by VisBench.