File size: 1,610 Bytes
362c76c | 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 | ---
license: apache-2.0
---
For details about the models, please see: https://github.com/roboflow/rf-detr
The models have been exported to executorch without lowering.
To run:
```python
from PIL import Image, ImageDraw
from executorch.runtime import Runtime
import torch
import torch.nn.functional as F
from torchvision import transforms
IMG_SIZE = (512, 512)
# change to (384, 384) for RFDETRNano
# change to (512, 512) for RFDETRSmall
# change to (576, 576) for RFDETRMedium
# change to (704, 704) for RFDETRLarge
def visualize_output(image, output):
draw = ImageDraw.Draw(image)
for box,logits in zip(output[0][0], output[1][0]):
probs = F.softmax(logits, dim=0)
pred_class = torch.argmax(probs, dim=0)
if probs[pred_class] > 0.7: # only draw if confidence is greater than 0.7
cx, cy, w, h = box
x1 = int((cx - w / 2) * img.width)
y1 = int((cy - h / 2) * img.height)
x2 = int((cx + w / 2) * img.width)
y2 = int((cy + h / 2) * img.height)
draw.rectangle([(x1, y1), (x2, y2)], fill=None, outline="black", width=3)
img = Image.open("./cats_coco.jpg").convert("RGB")
transform = transforms.Compose([
transforms.Resize(IMG_SIZE),
transforms.ToTensor(),
])
tensor = transform(img)
tensor = tensor.unsqueeze(0)
runtime = Runtime.get()
method = runtime.load_program("model_small.pte").load_method("forward")
outputs = method.execute([tensor])
visualize_output(img, outputs)
img.save("output.png")
img.show()
```
Example output:
|