File size: 1,557 Bytes
6994a4e 59c4610 6994a4e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 | ---
library_name: visbench
tags:
- visbench
- probing
- edge
---
# edge 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-edge-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 | `edge` (low_level) |
| pooling | `mean` (requested `mean`) |
| feature mode | `dense_only` |
| layers | `None` |
**Reported scores**
| metric | value |
| --- | --- |
| `edge_correlation` | 0.4558 |
| `mae` | 0.5028 |
| `rmse` | 0.9226 |
## Reproducing it
Fitted with:
- `activation`: `identity`
- `batch_size`: `8`
- `epochs`: `10`
- `head`: `linear`
- `hidden_dim`: `512`
- `layers`: `None`
- `loss`: `l1`
- `lr`: `0.0005`
- `optimizer`: `adamw`
- `protocol`: `visbench_edge_regression`
- `warmup_epochs`: `1.5`
- `weight_decay`: `0.0001`
Generated by [VisBench](https://github.com/turhancan97/VisBench).
|