---
base_model: PekingU/rtdetr_r101vd_coco_o365
datasets: keremberke/satellite-building-segmentation
library_name: transformers
license: mit
metrics:
- Average Precision (AP)
- Average Recall (AR)
pipeline_tag: object-detection
tags:
- remote sensing
- object detection
widget:
- src: img.png
  output:
    url: img.png
model-index:
- name: rt-detr-finetuned-for-satellite-image-roofs-detection
  results:
  - task:
      type: object-detection
    dataset:
      name: keremberke/satellite-building-segmentation
      type: image-segmentation
    metrics:
    - type: AP (IoU=0.50:0.95)
      value: 0.434
      name: AP @ IoU=0.50:0.95 | area=all | maxDets=100
    - type: AP (IoU=0.50)
      value: 0.652
      name: AP @ IoU=0.50 | area=all | maxDets=100
    - type: AP (IoU=0.75)
      value: 0.464
      name: AP @ IoU=0.75 | area=all | maxDets=100
    - type: AP (IoU=0.50:0.95) small objects
      value: 0.248
      name: AP @ IoU=0.50:0.95 | area=small | maxDets=100
    - type: AP (IoU=0.50:0.95) medium objects
      value: 0.510
      name: AP @ IoU=0.50:0.95 | area=medium | maxDets=100
    - type: AP (IoU=0.50:0.95) large objects
      value: 0.632
      name: AP @ IoU=0.50:0.95 | area=large | maxDets=100
    - type: AR (IoU=0.50:0.95) maxDets=1
      value: 0.056
      name: AR @ IoU=0.50:0.95 | area=all | maxDets=1
    - type: AR (IoU=0.50:0.95) maxDets=10
      value: 0.328
      name: AR @ IoU=0.50:0.95 | area=all | maxDets=10
    - type: AR (IoU=0.50:0.95) maxDets=100
      value: 0.519
      name: AR @ IoU=0.50:0.95 | area=all | maxDets=100
    - type: AR (IoU=0.50:0.95) small objects
      value: 0.337
      name: AR @ IoU=0.50:0.95 | area=small | maxDets=100
    - type: AR (IoU=0.50:0.95) medium objects
      value: 0.601
      name: AR @ IoU=0.50:0.95 | area=medium | maxDets=100
    - type: AR (IoU=0.50:0.95) large objects
      value: 0.714
      name: AR @ IoU=0.50:0.95 | area=large | maxDets=100
---

# Model Card

Roof Detection for Remote Sensing task.
<!-- Provide a quick summary of what the model is/does. -->



## Model Details

### Model Description

<!-- Provide a longer summary of what this model is. -->

- **Model type:** Object Detection for Remote Sensing task.
- **License:** MIT

### Model Sources 

<!-- Provide the basic links for the model. -->

- **GitHub:** [Jupyter Notebook](https://github.com/ownEyes/satellite-image-roofs-auto-annotation-sourcecode/blob/dev/notebooks/finetune_rtdetr.ipynb)
- **Demo:** [Hugging Face Space](https://huggingface.co/spaces/Yifeng-Liu/satellite-image-roofs-auto-annotation)



## Limitations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.

## How to Get Started with the Model

Use the code below to get started with the model.
```python
from transformers import AutoModelForObjectDetection, AutoImageProcessor
import torch
import cv2

image_path=YOUR_IMAGE_PATH
image = cv2.imread(image_path)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

model = AutoModelForObjectDetection.from_pretrained("Yifeng-Liu/rt-detr-finetuned-for-satellite-image-roofs-detection")
image_processor = AutoImageProcessor.from_pretrained("Yifeng-Liu/rt-detr-finetuned-for-satellite-image-roofs-detection")


CONFIDENCE_TRESHOLD = 0.5

with torch.no_grad():
    model.to(device)

    # load image and predict
    inputs = image_processor(images=image, return_tensors='pt').to(device)
    outputs = model(**inputs)

    # post-process
    target_sizes = torch.tensor([image.shape[:2]]).to(device)
    results = image_processor.post_process_object_detection(
        outputs=outputs,
        threshold=CONFIDENCE_TRESHOLD,
        target_sizes=target_sizes
    )[0]
```