Update README.md
Browse files
README.md
CHANGED
|
@@ -10,31 +10,26 @@ tags:
|
|
| 10 |
language:
|
| 11 |
- en
|
| 12 |
---
|
| 13 |
-
|
| 14 |
# Fashion Category Classifier
|
| 15 |
-
|
| 16 |
DistilBERT fine-tuned to classify Indian fashion product titles into 11 categories.
|
| 17 |
|
| 18 |
-
##
|
|
|
|
| 19 |
|
|
|
|
| 20 |
`Tops` `Bottoms` `Dresses` `Outerwear` `Footwear` `Bags` `Jewellery` `Ethnicwear` `Activewear` `Innerwear` `Headwear`
|
| 21 |
|
| 22 |
## Usage
|
| 23 |
-
|
| 24 |
```python
|
| 25 |
from transformers import pipeline
|
| 26 |
-
|
| 27 |
clf = pipeline("text-classification", model="roaringguts/DistilBERT")
|
| 28 |
-
|
| 29 |
clf("Nike Dry Fit Running Tshirt")
|
| 30 |
# [{'label': 'Activewear', 'score': 0.99}]
|
| 31 |
-
|
| 32 |
# batch
|
| 33 |
clf(["Lavie Women Tote Bag", "Malabar Gold Plated Necklace", "Clarks Oxford Shoes"])
|
| 34 |
```
|
| 35 |
|
| 36 |
## Training
|
| 37 |
-
|
| 38 |
- **Base model:** `distilbert-base-uncased`
|
| 39 |
- **Dataset:** ~63k real product titles from Myntra + synthetic samples generated for underrepresented categories
|
| 40 |
- **Split:** 80/10/10 train/val/test, stratified
|
|
@@ -44,7 +39,6 @@ clf(["Lavie Women Tote Bag", "Malabar Gold Plated Necklace", "Clarks Oxford Shoe
|
|
| 44 |
- **Precision:** fp16
|
| 45 |
|
| 46 |
## Evaluation
|
| 47 |
-
|
| 48 |
Evaluated on a held-out stratified test set.
|
| 49 |
|
| 50 |
| Metric | Score |
|
|
@@ -52,11 +46,9 @@ Evaluated on a held-out stratified test set.
|
|
| 52 |
| Accuracy | **99.27%** |
|
| 53 |
|
| 54 |
## Limitations
|
| 55 |
-
|
| 56 |
- Trained on Indian e-commerce titles — may underperform on Western brand naming conventions
|
| 57 |
- Ethnicwear and Innerwear have some overlap (e.g. sports bras vs regular bras)
|
| 58 |
- Low sample count for Bags and Headwear in original data, partially filled with synthetic titles
|
| 59 |
|
| 60 |
## License
|
| 61 |
-
|
| 62 |
-
[CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)
|
|
|
|
| 10 |
language:
|
| 11 |
- en
|
| 12 |
---
|
|
|
|
| 13 |
# Fashion Category Classifier
|
|
|
|
| 14 |
DistilBERT fine-tuned to classify Indian fashion product titles into 11 categories.
|
| 15 |
|
| 16 |
+
## Credits
|
| 17 |
+
Built on [DistilBERT](https://huggingface.co/distilbert/distilbert-base-uncased) by Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf (Hugging Face), licensed under [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0).
|
| 18 |
|
| 19 |
+
## Labels
|
| 20 |
`Tops` `Bottoms` `Dresses` `Outerwear` `Footwear` `Bags` `Jewellery` `Ethnicwear` `Activewear` `Innerwear` `Headwear`
|
| 21 |
|
| 22 |
## Usage
|
|
|
|
| 23 |
```python
|
| 24 |
from transformers import pipeline
|
|
|
|
| 25 |
clf = pipeline("text-classification", model="roaringguts/DistilBERT")
|
|
|
|
| 26 |
clf("Nike Dry Fit Running Tshirt")
|
| 27 |
# [{'label': 'Activewear', 'score': 0.99}]
|
|
|
|
| 28 |
# batch
|
| 29 |
clf(["Lavie Women Tote Bag", "Malabar Gold Plated Necklace", "Clarks Oxford Shoes"])
|
| 30 |
```
|
| 31 |
|
| 32 |
## Training
|
|
|
|
| 33 |
- **Base model:** `distilbert-base-uncased`
|
| 34 |
- **Dataset:** ~63k real product titles from Myntra + synthetic samples generated for underrepresented categories
|
| 35 |
- **Split:** 80/10/10 train/val/test, stratified
|
|
|
|
| 39 |
- **Precision:** fp16
|
| 40 |
|
| 41 |
## Evaluation
|
|
|
|
| 42 |
Evaluated on a held-out stratified test set.
|
| 43 |
|
| 44 |
| Metric | Score |
|
|
|
|
| 46 |
| Accuracy | **99.27%** |
|
| 47 |
|
| 48 |
## Limitations
|
|
|
|
| 49 |
- Trained on Indian e-commerce titles — may underperform on Western brand naming conventions
|
| 50 |
- Ethnicwear and Innerwear have some overlap (e.g. sports bras vs regular bras)
|
| 51 |
- Low sample count for Bags and Headwear in original data, partially filled with synthetic titles
|
| 52 |
|
| 53 |
## License
|
| 54 |
+
This model is released under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). The base DistilBERT weights it derives from are licensed under Apache 2.0.
|
|
|