README.md

---
pipeline_tag: image-classification
license: apache-2.0
base_model: timm/xception65.tf_in1k
library_name: zeromodels
tags:
- keras
- zeromodels
- image-classification
- xception
- backbone
- arxiv:1610.02357
- pytorch
- jax
- tf
---

## ***See [our collection](https://huggingface.co/collections/zeromodels/xception-6a8eae60db7ae3e5f3bf9d9a) for all versions of Xception.***

# Run Xception 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-Xception-blue)](https://imvision12.github.io/ZeroModels/classification_backbones/) [![Collection](https://img.shields.io/badge/HF-Xception%20collection-yellow)](https://huggingface.co/collections/zeromodels/xception-6a8eae60db7ae3e5f3bf9d9a)

# zeromodels/xception65_tf_in1k

Paper: [Xception: Deep Learning with Depthwise Separable Convolutions (arXiv:1610.02357)](https://arxiv.org/abs/1610.02357) · [HF Papers](https://huggingface.co/papers/1610.02357)

Xception interprets Inception modules as depthwise separable convolutions. Classifier or entry/middle/exit-flow backbone.

For more details on the model, please go to the upstream [model card](https://huggingface.co/timm/xception65.tf_in1k).

Pure-**Keras 3** conversion of [`timm/xception65.tf_in1k`](https://huggingface.co/timm/xception65.tf_in1k) for [zeromodels](https://github.com/IMvision12/ZeroModels). One implementation runs unmodified on **TensorFlow / Torch / JAX**.

This is an **image-classification / backbone** checkpoint (`XceptionImageClassify` / `XceptionModel`).

## ✨ Quick start

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

from PIL import Image
import numpy as np
from zeromodels.models.xception import XceptionImageClassify, XceptionModel

model = XceptionImageClassify.from_weights("zeromodels/xception65_tf_in1k")
backbone = XceptionModel.from_weights(
    "zeromodels/xception65_tf_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 Xception variant the same way with `from_weights("zeromodels/<variant>")`:

| Variant | Hub |
|---|---|
| `xception41_tf_in1k` | [`zeromodels/xception41_tf_in1k`](https://huggingface.co/zeromodels/xception41_tf_in1k) |
| `xception41p_ra3_in1k` | [`zeromodels/xception41p_ra3_in1k`](https://huggingface.co/zeromodels/xception41p_ra3_in1k) |
| `xception65_ra3_in1k` | [`zeromodels/xception65_ra3_in1k`](https://huggingface.co/zeromodels/xception65_ra3_in1k) |
| `xception65_tf_in1k` | [`zeromodels/xception65_tf_in1k`](https://huggingface.co/zeromodels/xception65_tf_in1k) |
| `xception65p_ra3_in1k` | [`zeromodels/xception65p_ra3_in1k`](https://huggingface.co/zeromodels/xception65p_ra3_in1k) |
| `xception71_tf_in1k` | [`zeromodels/xception71_tf_in1k`](https://huggingface.co/zeromodels/xception71_tf_in1k) |

## Tips

- Set `KERAS_BACKEND` **before** importing Keras / zeromodels.
- `XceptionImageClassify` returns class logits; `XceptionModel` 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: `XceptionImageClassify.from_weights("hf:timm/xception65.tf_in1k")`.

## Special Thanks

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

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