You can deploy this model and start earning money today!

turhancan97/
visbench-depth-dinov2_vits14

Name: visbench-depth-dinov2_vits14
Availability: InStock

visbench

probing

depth

Views

No views yet

depth 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.

python

import visbench
from visbench.hub import load_probe_from_hub

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

It only works with this backbone

These weights were fitted on features from dinov2_vits14, taken with pooling=mean 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	`depth` (mid_level)
pooling	`mean` (requested `mean`)
feature mode	`dense_only`
layers	`None`

Reported scores

metric	value
`abs_rel`	0.1639
`d1`	0.7652
`d2`	0.9593
`d3`	0.9931
`rmse`	0.5518

Reproducing it

Fitted with:

batch_size: 8
epochs: 10
head: linear
hidden_dim: 512
layers: None
lr: 0.0005
max_depth: 10.0
min_depth: 0.001
n_bins: 256
optimizer: adamw
protocol: probe3d
scale_invariant: False
warmup_epochs: 1.5
weight_decay: 0.0001

Generated by VisBench.

turhancan97/visbench-depth-dinov2_vits14

depth probe for dinov2_vits14

It only works with this backbone

Reproducing it

turhancan97/
visbench-depth-dinov2_vits14

depth probe for `dinov2_vits14`