| --- |
| library_name: visbench |
| tags: |
| - visbench |
| - probing |
| - detection |
| --- |
| |
| # detection probe for `dinov2_vitb14` |
| |
| A **trained probe head**, not a backbone. It is the small module VisBench fits |
| on top of frozen `dinov2_vitb14` features to measure what those features carry. |
|
|
| ```python |
| import visbench |
| from visbench.hub import load_probe_from_hub |
| |
| backbone = visbench.get_backbone("dinov2_vitb14") |
| probe = load_probe_from_hub("turhancan97/visbench-detection-dinov2_vitb14", backbone=backbone) |
| ``` |
|
|
| ## It only works with this backbone |
|
|
| These weights were fitted on features from `dinov2_vitb14`, 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_vitb14` | |
| | backbone key | `dinov2/dinov2_vitb14/224/7764ea0f912e` | |
| | task | `detection` (high_level) | |
| | pooling | `mean` (requested `mean`) | |
| | feature mode | `dense_only` | |
| | layers | `None` | |
|
|
| **Reported scores** |
|
|
| | metric | value | |
| | --- | --- | |
| | `classes_scored` | 20.0000 | |
| | `detections_per_image` | 88.5717 | |
| | `map_50` | 0.2897 | |
| | `map_50_95` | 0.0988 | |
|
|
| ## Reproducing it |
|
|
| Fitted with: |
|
|
| - `batch_size`: `8` |
| - `box_weight`: `2.0` |
| - `epochs`: `10` |
| - `focal_alpha`: `0.25` |
| - `focal_gamma`: `2.0` |
| - `head`: `detection` |
| - `hidden_dim`: `0` |
| - `image_size`: `224` |
| - `iou_thresholds`: `[0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95]` |
| - `lr`: `0.0005` |
| - `max_detections`: `100` |
| - `nms_iou`: `0.5` |
| - `num_classes`: `20` |
| - `optimizer`: `adamw` |
| - `protocol`: `visbench_anchor_free_det` |
| - `score_threshold`: `0.05` |
| - `warmup_epochs`: `1.5` |
| - `weight_decay`: `0.0001` |
|
|
| Generated by [VisBench](https://github.com/turhancan97/VisBench). |
|
|
