|
|
---
|
|
|
license: apache-2.0
|
|
|
datasets:
|
|
|
- frgfm/imagenette
|
|
|
language:
|
|
|
- en
|
|
|
base_model:
|
|
|
- google/siglip2-base-patch16-224
|
|
|
pipeline_tag: image-classification
|
|
|
library_name: transformers
|
|
|
tags:
|
|
|
- ImageNet
|
|
|
- SigLIP2
|
|
|
- Classifier
|
|
|
---
|
|
|
|
|
|
 |
|
|
|
|
|
# IMAGENETTE |
|
|
|
|
|
> IMAGENETTE is a vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for multi-class image classification. It is trained to classify images into 10 categories from the popular Imagenette dataset using the SiglipForImageClassification architecture. |
|
|
|
|
|
> [!note] |
|
|
*SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features* https://arxiv.org/pdf/2502.14786 |
|
|
|
|
|
> [!note] |
|
|
> *ImageNet Large Scale Visual Recognition Challenge* https://arxiv.org/pdf/1409.0575 |
|
|
|
|
|
```py |
|
|
Classification Report: |
|
|
precision recall f1-score support |
|
|
|
|
|
tench 0.9885 0.9834 0.9859 963 |
|
|
english springer 0.9843 0.9822 0.9832 955 |
|
|
cassette player 0.9544 0.9486 0.9515 993 |
|
|
chain saw 0.9257 0.8998 0.9125 858 |
|
|
church 0.9654 0.9798 0.9726 941 |
|
|
French horn 0.9757 0.9665 0.9711 956 |
|
|
garbage truck 0.8883 0.9761 0.9301 961 |
|
|
gas pump 0.9366 0.9044 0.9202 931 |
|
|
golf ball 0.9925 0.9716 0.9819 951 |
|
|
parachute 0.9821 0.9708 0.9764 960 |
|
|
|
|
|
accuracy 0.9590 9469 |
|
|
macro avg 0.9593 0.9583 0.9586 9469 |
|
|
weighted avg 0.9597 0.9590 0.9591 9469 |
|
|
``` |
|
|
|
|
|
 |
|
|
|
|
|
--- |
|
|
|
|
|
## Label Space: 10 Classes |
|
|
|
|
|
The model predicts one of the following image classes: |
|
|
|
|
|
``` |
|
|
0: tench |
|
|
1: english springer |
|
|
2: cassette player |
|
|
3: chain saw |
|
|
4: church |
|
|
5: French horn |
|
|
6: garbage truck |
|
|
7: gas pump |
|
|
8: golf ball |
|
|
9: parachute |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Install Dependencies |
|
|
|
|
|
```bash |
|
|
pip install -q transformers torch pillow gradio hf_xet |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Inference Code |
|
|
|
|
|
```python |
|
|
import gradio as gr |
|
|
from transformers import AutoImageProcessor, SiglipForImageClassification |
|
|
from PIL import Image |
|
|
import torch |
|
|
|
|
|
# Load model and processor |
|
|
model_name = "prithivMLmods/IMAGENETTE" |
|
|
model = SiglipForImageClassification.from_pretrained(model_name) |
|
|
processor = AutoImageProcessor.from_pretrained(model_name) |
|
|
|
|
|
# Label mapping |
|
|
id2label = { |
|
|
"0": "tench", |
|
|
"1": "english springer", |
|
|
"2": "cassette player", |
|
|
"3": "chain saw", |
|
|
"4": "church", |
|
|
"5": "French horn", |
|
|
"6": "garbage truck", |
|
|
"7": "gas pump", |
|
|
"8": "golf ball", |
|
|
"9": "parachute" |
|
|
} |
|
|
|
|
|
def classify_image(image): |
|
|
image = Image.fromarray(image).convert("RGB") |
|
|
inputs = processor(images=image, return_tensors="pt") |
|
|
|
|
|
with torch.no_grad(): |
|
|
outputs = model(**inputs) |
|
|
logits = outputs.logits |
|
|
probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist() |
|
|
|
|
|
prediction = { |
|
|
id2label[str(i)]: round(probs[i], 3) for i in range(len(probs)) |
|
|
} |
|
|
|
|
|
return prediction |
|
|
|
|
|
# Gradio Interface |
|
|
iface = gr.Interface( |
|
|
fn=classify_image, |
|
|
inputs=gr.Image(type="numpy"), |
|
|
outputs=gr.Label(num_top_classes=3, label="Image Classification"), |
|
|
title="IMAGENETTE - SigLIP2 Classifier", |
|
|
description="Upload an image to classify it into one of 10 categories from the Imagenette dataset." |
|
|
) |
|
|
|
|
|
if __name__ == "__main__": |
|
|
iface.launch() |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Intended Use |
|
|
|
|
|
IMAGENETTE is designed for: |
|
|
|
|
|
* Educational purposes and model benchmarking. |
|
|
* Demonstrating the performance of SigLIP2 on a small but diverse classification task. |
|
|
* Fine-tuning workflows on vision-language models. |
