Hugging Face's logo

zeromodels
/
vit_base_patch16_224_augreg_in1k

Image Classification
Keras
PyTorch
JAX
TensorFlow
zeromodels
vit
backbone
Model card Files
xet
 Community
IMvision12 commited on
Commit
6da158a
·
verified ·
1 Parent(s): d564b56

fix readme.md

Browse files
Files changed (1) hide show
  1. README.md +81 -6
README.md CHANGED
@@ -8,20 +8,95 @@ tags:
8
  - kerasformers
9
  - image-classification
10
  - vit
 
 
11
  - pytorch
12
  - jax
13
  - tf
14
  ---
15
 
16
- # vit_base_patch16_224_augreg_in1k
17
 
18
- Pure-Keras 3 image-classification weight for [kerasformers](https://github.com/IMvision12/KerasFormers), converted from [timm/vit_base_patch16_224.augreg_in1k](https://huggingface.co/timm/vit_base_patch16_224.augreg_in1k).
19
 
20
- ## Usage
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21
 
22
  ```python
23
- from kerasformers.models.vit import ViTImageClassify
24
- model = ViTImageClassify.from_weights("vit_base_patch16_224_augreg_in1k")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25
  ```
26
 
27
- License: **apache-2.0**, inherited from the upstream source.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8
  - kerasformers
9
  - image-classification
10
  - vit
11
+ - backbone
12
+ - arxiv:2010.11929
13
  - pytorch
14
  - jax
15
  - tf
16
  ---
17
 
18
+ ## ***See [our collection](https://huggingface.co/collections/kerasformers/vit-6a6d51a16516088e6ab2ec60) for all versions of ViT.***
19
 
20
+ # Run ViT with Keras 3: JAX, PyTorch, or TensorFlow
21
 
22
+ [![GitHub](https://img.shields.io/badge/GitHub-KerasFormers-black?logo=github)](https://github.com/IMvision12/KerasFormers) [![Docs](https://img.shields.io/badge/Docs-ViT-blue)](https://imvision12.github.io/KerasFormers/classification_backbones/) [![Collection](https://img.shields.io/badge/HF-ViT%20collection-yellow)](https://huggingface.co/collections/kerasformers/vit-6a6d51a16516088e6ab2ec60)
23
+
24
+ # kerasformers/vit_base_patch16_224_augreg_in1k
25
+
26
+ Paper: [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale (arXiv:2010.11929)](https://arxiv.org/abs/2010.11929) · [HF Papers](https://huggingface.co/papers/2010.11929)
27
+
28
+ Vision Transformer (ViT) patches an image and runs a transformer encoder. Use `ViTImageClassify` for logits or `ViTModel` for tokens / per-block features via `as_backbone=True`.
29
+
30
+ For more details on the model, please go to the upstream [model card](https://huggingface.co/timm/vit_base_patch16_224.augreg_in1k).
31
+
32
+ Pure-**Keras 3** conversion of [`timm/vit_base_patch16_224.augreg_in1k`](https://huggingface.co/timm/vit_base_patch16_224.augreg_in1k) for [kerasformers](https://github.com/IMvision12/KerasFormers). One implementation runs unmodified on **TensorFlow / Torch / JAX**.
33
+
34
+ This is an **image-classification / backbone** checkpoint (`ViTImageClassify` / `ViTModel`).
35
+
36
+ ## ✨ Quick start
37
 
38
  ```python
39
+ import os
40
+ os.environ["KERAS_BACKEND"] = "torch" # or "jax" / "tensorflow"
41
+
42
+ from PIL import Image
43
+ import numpy as np
44
+ from kerasformers.models.vit import ViTImageClassify, ViTModel
45
+
46
+ model = ViTImageClassify.from_weights("kerasformers/vit_base_patch16_224_augreg_in1k")
47
+ backbone = ViTModel.from_weights(
48
+ "kerasformers/vit_base_patch16_224_augreg_in1k", as_backbone=True
49
+ )
50
+
51
+ image = Image.open("your_image.jpg").convert("RGB")
52
+ image = image.resize((224, 224))
53
+ x = np.asarray(image, dtype="float32")[None] # (1, H, W, 3)
54
+ print(model(x).shape) # (1, num_classes)
55
+ feats = backbone(x)
56
+ print(len(feats), [tuple(f.shape) for f in feats])
57
  ```
58
 
59
+ Load any ViT variant the same way with `from_weights("kerasformers/<variant>")`:
60
+
61
+ | Variant | Hub |
62
+ |---|---|
63
+ | `vit_base_patch16_224_augreg_in1k` | [`kerasformers/vit_base_patch16_224_augreg_in1k`](https://huggingface.co/kerasformers/vit_base_patch16_224_augreg_in1k) |
64
+ | `vit_base_patch16_224_augreg_in21k` | [`kerasformers/vit_base_patch16_224_augreg_in21k`](https://huggingface.co/kerasformers/vit_base_patch16_224_augreg_in21k) |
65
+ | `vit_base_patch16_224_augreg_in21k_ft_in1k` | [`kerasformers/vit_base_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_base_patch16_224_augreg_in21k_ft_in1k) |
66
+ | `vit_base_patch16_224_orig_in21k_ft_in1k` | [`kerasformers/vit_base_patch16_224_orig_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_base_patch16_224_orig_in21k_ft_in1k) |
67
+ | `vit_base_patch16_384_augreg_in1k` | [`kerasformers/vit_base_patch16_384_augreg_in1k`](https://huggingface.co/kerasformers/vit_base_patch16_384_augreg_in1k) |
68
+ | `vit_base_patch16_384_augreg_in21k_ft_in1k` | [`kerasformers/vit_base_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_base_patch16_384_augreg_in21k_ft_in1k) |
69
+ | `vit_base_patch16_384_orig_in21k_ft_in1k` | [`kerasformers/vit_base_patch16_384_orig_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_base_patch16_384_orig_in21k_ft_in1k) |
70
+ | `vit_base_patch32_224_augreg_in1k` | [`kerasformers/vit_base_patch32_224_augreg_in1k`](https://huggingface.co/kerasformers/vit_base_patch32_224_augreg_in1k) |
71
+ | `vit_base_patch32_224_augreg_in21k` | [`kerasformers/vit_base_patch32_224_augreg_in21k`](https://huggingface.co/kerasformers/vit_base_patch32_224_augreg_in21k) |
72
+ | `vit_base_patch32_224_augreg_in21k_ft_in1k` | [`kerasformers/vit_base_patch32_224_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_base_patch32_224_augreg_in21k_ft_in1k) |
73
+ | `vit_base_patch32_384_augreg_in1k` | [`kerasformers/vit_base_patch32_384_augreg_in1k`](https://huggingface.co/kerasformers/vit_base_patch32_384_augreg_in1k) |
74
+ | `vit_base_patch32_384_augreg_in21k_ft_in1k` | [`kerasformers/vit_base_patch32_384_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_base_patch32_384_augreg_in21k_ft_in1k) |
75
+ | `vit_large_patch16_224_augreg_in21k` | [`kerasformers/vit_large_patch16_224_augreg_in21k`](https://huggingface.co/kerasformers/vit_large_patch16_224_augreg_in21k) |
76
+ | `vit_large_patch16_224_augreg_in21k_ft_in1k` | [`kerasformers/vit_large_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_large_patch16_224_augreg_in21k_ft_in1k) |
77
+ | `vit_large_patch16_384_augreg_in21k_ft_in1k` | [`kerasformers/vit_large_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_large_patch16_384_augreg_in21k_ft_in1k) |
78
+ | `vit_large_patch32_384_orig_in21k_ft_in1k` | [`kerasformers/vit_large_patch32_384_orig_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_large_patch32_384_orig_in21k_ft_in1k) |
79
+ | `vit_small_patch16_224_augreg_in1k` | [`kerasformers/vit_small_patch16_224_augreg_in1k`](https://huggingface.co/kerasformers/vit_small_patch16_224_augreg_in1k) |
80
+ | `vit_small_patch16_224_augreg_in21k` | [`kerasformers/vit_small_patch16_224_augreg_in21k`](https://huggingface.co/kerasformers/vit_small_patch16_224_augreg_in21k) |
81
+ | `vit_small_patch16_224_augreg_in21k_ft_in1k` | [`kerasformers/vit_small_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_small_patch16_224_augreg_in21k_ft_in1k) |
82
+ | `vit_small_patch16_384_augreg_in1k` | [`kerasformers/vit_small_patch16_384_augreg_in1k`](https://huggingface.co/kerasformers/vit_small_patch16_384_augreg_in1k) |
83
+ | `vit_small_patch16_384_augreg_in21k_ft_in1k` | [`kerasformers/vit_small_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_small_patch16_384_augreg_in21k_ft_in1k) |
84
+ | `vit_small_patch32_224_augreg_in21k` | [`kerasformers/vit_small_patch32_224_augreg_in21k`](https://huggingface.co/kerasformers/vit_small_patch32_224_augreg_in21k) |
85
+ | `vit_small_patch32_224_augreg_in21k_ft_in1k` | [`kerasformers/vit_small_patch32_224_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_small_patch32_224_augreg_in21k_ft_in1k) |
86
+ | `vit_small_patch32_384_augreg_in21k_ft_in1k` | [`kerasformers/vit_small_patch32_384_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_small_patch32_384_augreg_in21k_ft_in1k) |
87
+ | `vit_tiny_patch16_224_augreg_in21k` | [`kerasformers/vit_tiny_patch16_224_augreg_in21k`](https://huggingface.co/kerasformers/vit_tiny_patch16_224_augreg_in21k) |
88
+ | `vit_tiny_patch16_224_augreg_in21k_ft_in1k` | [`kerasformers/vit_tiny_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_tiny_patch16_224_augreg_in21k_ft_in1k) |
89
+ | `vit_tiny_patch16_384_augreg_in21k_ft_in1k` | [`kerasformers/vit_tiny_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/kerasformers/vit_tiny_patch16_384_augreg_in21k_ft_in1k) |
90
+
91
+ ## Tips
92
+
93
+ - Set `KERAS_BACKEND` **before** importing Keras / kerasformers.
94
+ - `ViTImageClassify` returns class logits; `ViTModel` returns features (`as_backbone=True` for multi-scale stages).
95
+ - See [docs](https://imvision12.github.io/KerasFormers/classification_backbones/) and [Loading Weights](https://imvision12.github.io/KerasFormers/loading_weights/).
96
+ - Upstream / timm checkpoints: `ViTImageClassify.from_weights("hf:timm/vit_base_patch16_224.augreg_in1k")`.
97
+
98
+ ## Special Thanks
99
+
100
+ A huge thank you to the ViT authors and the timm / Hub communities for creating and releasing these models.
101
+
102
+ License: see YAML `license` (usually matches the upstream checkpoint).