README.md

---
pipeline_tag: image-classification
license: apache-2.0
base_model: timm/inception_next_tiny.sail_in1k
library_name: zeromodels
tags:
- keras
- zeromodels
- image-classification
- inception-next
- backbone
- arxiv:2303.16900
- pytorch
- jax
- tf
---

## ***See [our collection](https://huggingface.co/collections/zeromodels/inceptionnext-6a8eaeef3fde5bbc3359a23f) for all versions of InceptionNeXt.***

# Run InceptionNeXt with Keras 3: JAX, PyTorch, or TensorFlow

[![GitHub](https://img.shields.io/badge/GitHub-ZeroModels-black?logo=github)](https://github.com/IMvision12/ZeroModels) [![Docs](https://img.shields.io/badge/Docs-InceptionNeXt-blue)](https://imvision12.github.io/ZeroModels/classification_backbones/) [![Collection](https://img.shields.io/badge/HF-InceptionNeXt%20collection-yellow)](https://huggingface.co/collections/zeromodels/inceptionnext-6a8eaeef3fde5bbc3359a23f)

# zeromodels/inception_next_tiny_sail_in1k

Paper: [InceptionNeXt: When Inception Meets ConvNeXt (arXiv:2303.16900)](https://arxiv.org/abs/2303.16900) · [HF Papers](https://huggingface.co/papers/2303.16900)

InceptionNeXt blends Inception-style token mixing with a ConvNeXt meta-architecture for efficient ImageNet classification.

For more details on the model, please go to the upstream [model card](https://huggingface.co/timm/inception_next_tiny.sail_in1k).

Pure-**Keras 3** conversion of [`timm/inception_next_tiny.sail_in1k`](https://huggingface.co/timm/inception_next_tiny.sail_in1k) for [zeromodels](https://github.com/IMvision12/ZeroModels). One implementation runs unmodified on **TensorFlow / Torch / JAX**.

This is an **image-classification / backbone** checkpoint (`InceptionNextImageClassify` / `InceptionNextModel`).

## ✨ Quick start

```python
import os
os.environ["KERAS_BACKEND"] = "torch"  # or "jax" / "tensorflow"

from PIL import Image
import numpy as np
from zeromodels.models.inception_next import InceptionNextImageClassify, InceptionNextModel

model = InceptionNextImageClassify.from_weights("zeromodels/inception_next_tiny_sail_in1k")
backbone = InceptionNextModel.from_weights(
    "zeromodels/inception_next_tiny_sail_in1k", as_backbone=True
)

image = Image.open("your_image.jpg").convert("RGB")
image = image.resize((224, 224))
x = np.asarray(image, dtype="float32")[None]  # (1, H, W, 3)
print(model(x).shape)  # (1, num_classes)
feats = backbone(x)
print(len(feats), [tuple(f.shape) for f in feats])
```

Load any InceptionNeXt variant the same way with `from_weights("zeromodels/<variant>")`:

| Variant | Hub |
|---|---|
| `inception_next_atto_sail_in1k` | [`zeromodels/inception_next_atto_sail_in1k`](https://huggingface.co/zeromodels/inception_next_atto_sail_in1k) |
| `inception_next_base_sail_in1k` | [`zeromodels/inception_next_base_sail_in1k`](https://huggingface.co/zeromodels/inception_next_base_sail_in1k) |
| `inception_next_base_sail_in1k_384` | [`zeromodels/inception_next_base_sail_in1k_384`](https://huggingface.co/zeromodels/inception_next_base_sail_in1k_384) |
| `inception_next_small_sail_in1k` | [`zeromodels/inception_next_small_sail_in1k`](https://huggingface.co/zeromodels/inception_next_small_sail_in1k) |
| `inception_next_tiny_sail_in1k` | [`zeromodels/inception_next_tiny_sail_in1k`](https://huggingface.co/zeromodels/inception_next_tiny_sail_in1k) |

## Tips

- Set `KERAS_BACKEND` **before** importing Keras / zeromodels.
- `InceptionNextImageClassify` returns class logits; `InceptionNextModel` returns features (`as_backbone=True` for multi-scale stages).
- See [docs](https://imvision12.github.io/ZeroModels/classification_backbones/) and [Loading Weights](https://imvision12.github.io/ZeroModels/loading_weights/).
- Upstream / timm checkpoints: `InceptionNextImageClassify.from_weights("hf:timm/inception_next_tiny.sail_in1k")`.

## Special Thanks

A huge thank you to the InceptionNeXt authors and the timm / Hub communities for creating and releasing these models.

License: see YAML `license` (usually matches the upstream checkpoint).