Image Segmentation
Transformers
Safetensors
inkdetection_resnet3d
feature-extraction
vesuvius-challenge
ink-detection
herculaneum
resnet3d
u-net
3d-segmentation
volumetric-imaging
custom_code
Instructions to use scrollprize/PHerc.1667-iteration-1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use scrollprize/PHerc.1667-iteration-1 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-segmentation", model="scrollprize/PHerc.1667-iteration-1", trust_remote_code=True)# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("scrollprize/PHerc.1667-iteration-1", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
Point W&B link at canonical run URL (/runs/<id>)
Browse files
README.md
CHANGED
|
@@ -136,7 +136,7 @@ cv2.imwrite("prediction.png", np.clip(pred * 255, 0, 255).astype(np.uint8))
|
|
| 136 |
| **Training tiles** (256×256 sub-tiles at stride 64) | **3,396** |
|
| 137 |
| **Final train loss (`_epoch`)** | **0.4219** |
|
| 138 |
| **Final train loss (`_step`, single-batch noise)** | 0.4381 |
|
| 139 |
-
| **Wandb** | [vesuvius-challenge/paper/l2_ink1_l5infer](https://wandb.ai/vesuvius-challenge/paper/
|
| 140 |
| **Random seed** | 130697 |
|
| 141 |
| **Determinism** | `cudnn.deterministic = True`, `cudnn.benchmark = False` |
|
| 142 |
| **Hardware** | 1 × NVIDIA H100 80 GB; ≈ 2 h end-to-end (load + train + inference) |
|
|
|
|
| 136 |
| **Training tiles** (256×256 sub-tiles at stride 64) | **3,396** |
|
| 137 |
| **Final train loss (`_epoch`)** | **0.4219** |
|
| 138 |
| **Final train loss (`_step`, single-batch noise)** | 0.4381 |
|
| 139 |
+
| **Wandb** | [vesuvius-challenge/paper/l2_ink1_l5infer](https://wandb.ai/vesuvius-challenge/paper/runs/xlj01hgh) |
|
| 140 |
| **Random seed** | 130697 |
|
| 141 |
| **Determinism** | `cudnn.deterministic = True`, `cudnn.benchmark = False` |
|
| 142 |
| **Hardware** | 1 × NVIDIA H100 80 GB; ≈ 2 h end-to-end (load + train + inference) |
|