Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- anomaly-detection
|
| 4 |
+
- 3d-printing
|
| 5 |
+
- patchcore
|
| 6 |
+
- onnx
|
| 7 |
+
---
|
| 8 |
+
# 3D Print Failure Detector — PatchCore
|
| 9 |
+
|
| 10 |
+
Anomaly detection model for detecting 3D print failures (spaghetti, detachment, etc.).
|
| 11 |
+
Trained **only on normal prints** using PatchCore with `resnet18` backbone.
|
| 12 |
+
|
| 13 |
+
## Metrics
|
| 14 |
+
```
|
| 15 |
+
{
|
| 16 |
+
"image_AUROC": 0.8632,
|
| 17 |
+
"image_F1Score": 0.9248
|
| 18 |
+
}
|
| 19 |
+
```
|
| 20 |
+
|
| 21 |
+
## Threshold
|
| 22 |
+
Auto-computed threshold: `17.3327`
|
| 23 |
+
Frames with anomaly score > threshold are classified as **failure**.
|
| 24 |
+
|
| 25 |
+
## Dataset
|
| 26 |
+
- Train: 1491 normal frames from YouTube printer timelapse videos
|
| 27 |
+
- Test: 373 normal + 850 failure frames
|
| 28 |
+
|
| 29 |
+
## Inference
|
| 30 |
+
ONNX model: `weights/onnx/model.onnx`
|
| 31 |
+
Threshold config: `weights/onnx/config.json`
|