File size: 5,214 Bytes
e07530d a1aaf83 e07530d a1aaf83 e07530d a1aaf83 e07530d a1aaf83 e07530d a1aaf83 e07530d a1aaf83 4cb7dd8 a1aaf83 e07530d a1aaf83 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 | ---
pipeline_tag: zero-shot-object-detection
license: apache-2.0
base_model: google/owlv2-base-patch16-ensemble
library_name: kerasformers
tags:
- keras
- kerasformers
- owlv2
- open-vocabulary
- zero-shot-object-detection
- object-detection
- arxiv:2306.09683
- pytorch
- jax
- tf
---
## ***See [our collection](https://huggingface.co/collections/kerasformers/owlv2-6a6a7aaaf7cd6616646d8318) for all versions of OWLv2.***
# Run OWLv2 with Keras 3: JAX, PyTorch, or TensorFlow
[](https://github.com/IMvision12/KerasFormers) [](https://imvision12.github.io/KerasFormers/owlv2/) [](https://huggingface.co/collections/kerasformers/owlv2-6a6a7aaaf7cd6616646d8318)
# kerasformers/owlv2-base-patch16-ensemble
Paper: [Scaling Open-Vocabulary Object Detection (arXiv:2306.09683)](https://arxiv.org/abs/2306.09683) · [HF Papers](https://huggingface.co/papers/2306.09683)
OWLv2 keeps OWL-ViT's dual-tower skeleton and per-patch detection head, and scales it with self-training on web image-text pairs. It adds an objectness head (a learned is-this-patch-an-object score) and pads images to a square before resizing, which matters for post-processing target sizes.
For more details on the model, please go to Google's original [model card](https://huggingface.co/google/owlv2-base-patch16-ensemble).
Pure-**Keras 3** conversion of [`google/owlv2-base-patch16-ensemble`](https://huggingface.co/google/owlv2-base-patch16-ensemble) for [kerasformers](https://github.com/IMvision12/KerasFormers). One implementation runs unmodified on **TensorFlow / Torch / JAX**.
This is an **open-vocabulary object detection** checkpoint (`Owlv2Detect`): pass free-text prompts at inference time.
## ✨ Quick start
```python
import os
os.environ["KERAS_BACKEND"] = "torch" # or "jax" / "tensorflow"
from PIL import Image
from kerasformers.models.owlv2 import (
Owlv2Detect,
Owlv2Processor,
Owlv2ImageProcessor,
)
model = Owlv2Detect.from_weights("kerasformers/owlv2-base-patch16-ensemble")
processor = Owlv2Processor.from_weights("kerasformers/owlv2-base-patch16-ensemble")
image_processor = Owlv2ImageProcessor.from_weights("kerasformers/owlv2-base-patch16-ensemble")
image = Image.open("your_image.jpg").convert("RGB")
prompts = ["a photo of a mug", "a photo of a knife"]
inputs = processor(text=[prompts], images=image)
output = model(
{
"input_ids": inputs["input_ids"],
"pixel_values": inputs["pixel_values"],
}
)
results = image_processor.post_process_object_detection(
output,
threshold=0.1,
target_sizes=[(image.height, image.width)],
text_labels=[prompts],
)[0]
for score, name, box in zip(
results["scores"], results["text_labels"], results["boxes"]
):
print(f"{name}: {float(score):.3f} {box}")
```
Load any OWL-ViT / OWLv2 variant the same way with `from_weights("kerasformers/<variant>")` (use `Owlv2Detect` for this repo):
| Variant | Hub | Family |
|---|---|---|
| `owlvit-base-patch32` | [`kerasformers/owlvit-base-patch32`](https://huggingface.co/kerasformers/owlvit-base-patch32) | OWL-ViT |
| `owlvit-base-patch16` | [`kerasformers/owlvit-base-patch16`](https://huggingface.co/kerasformers/owlvit-base-patch16) | OWL-ViT |
| `owlvit-large-patch14` | [`kerasformers/owlvit-large-patch14`](https://huggingface.co/kerasformers/owlvit-large-patch14) | OWL-ViT |
| `owlv2-base-patch16` | [`kerasformers/owlv2-base-patch16`](https://huggingface.co/kerasformers/owlv2-base-patch16) | OWLv2 |
| `owlv2-base-patch16-ensemble` | [`kerasformers/owlv2-base-patch16-ensemble`](https://huggingface.co/kerasformers/owlv2-base-patch16-ensemble) | OWLv2 |
| `owlv2-base-patch16-finetuned` | [`kerasformers/owlv2-base-patch16-finetuned`](https://huggingface.co/kerasformers/owlv2-base-patch16-finetuned) | OWLv2 |
| `owlv2-large-patch14` | [`kerasformers/owlv2-large-patch14`](https://huggingface.co/kerasformers/owlv2-large-patch14) | OWLv2 |
| `owlv2-large-patch14-ensemble` | [`kerasformers/owlv2-large-patch14-ensemble`](https://huggingface.co/kerasformers/owlv2-large-patch14-ensemble) | OWLv2 |
| `owlv2-large-patch14-finetuned` | [`kerasformers/owlv2-large-patch14-finetuned`](https://huggingface.co/kerasformers/owlv2-large-patch14-finetuned) | OWLv2 |
## Tips
- Set `KERAS_BACKEND` **before** importing Keras / kerasformers.
- Prefer `Processor.from_weights(...)` so image size matches the variant.
- Open-vocab thresholds are often much lower than closed-set detectors (try `0.1`).
- OWLv2 pads to square before resize; pass the original `(height, width)` as `target_sizes` carefully (see the OWLv2 docs for the padding trap).
- See [OWLv2 docs](https://imvision12.github.io/KerasFormers/owlv2/) and [Loading Weights](https://imvision12.github.io/KerasFormers/loading_weights/).
- Community / upstream safetensors still work via the `hf:` prefix, e.g. `Owlv2Detect.from_weights("hf:google/owlv2-base-patch16-ensemble")`.
## Special Thanks
A huge thank you to the Google OWLv2 authors for creating and releasing these models.
License: Apache 2.0.
|