Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
base_model:
|
| 4 |
+
- timm/eva02_base_patch16_clip_224.merged2b_s8b_b131k
|
| 5 |
+
pipeline_tag: zero-shot-image-classification
|
| 6 |
+
library_name: open_clip
|
| 7 |
+
---
|
| 8 |
+
# anime-style-tag-clip
|
| 9 |
+
This model weight is based on eva02_base_patch16_clip.
|
| 10 |
+
The dataset size is small, which may result in slightly reduced generalization performance of the model.
|
| 11 |
+
|
| 12 |
+
# Dataset
|
| 13 |
+
I trained this model with anime-style image-tag data.
|
| 14 |
+
Train: 29187
|
| 15 |
+
Valid: 3000
|
| 16 |
+
|
| 17 |
+
# Validation results
|
| 18 |
+
R@1: 0.87733
|
| 19 |
+
R@5: 0.96767
|
| 20 |
+
R@10: 0.98233
|
| 21 |
+
|
| 22 |
+
# Training & Inference code
|
| 23 |
+
https://github.com/dudcjs2779/tag-clip.git
|