object_detection.py

from transformers import DetrImageProcessor, DetrForObjectDetection
import torch
from PIL import Image, ImageDraw, ImageFont
import gradio as gr


def render_result_in_image(image):
    """
    Render detected objects in the input image.

    Args:
        image (PIL.Image): Input image.

    Returns:
        PIL.Image: Image with bounding boxes and labels drawn.
    """
    processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50", revision="no_timm")
    model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50", revision="no_timm")

    inputs = processor(images=image, return_tensors="pt")
    outputs = model(**inputs)

    # convert outputs (bounding boxes and class logits) to COCO API
    # let's only keep detections with score > 0.9
    target_sizes = torch.tensor([image.size[::-1]])
    results = processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=0.9)[0]

    # Create id to label mapping
    id2label = {idx: model.config.id2label[idx] for idx in range(len(model.config.id2label))}

    # Render results in the image
    rendered_image = render_result_in_image_helper(image.copy(), results, id2label)
    return rendered_image


def render_result_in_image_helper(image, results, id2label):
    """
    Helper function to render detected objects in the input image.

    Args:
        image (PIL.Image): Input image.
        results (dict): Detection results containing 'scores', 'labels', and 'boxes'.
        id2label (dict): Mapping from class indices to class labels.

    Returns:
        PIL.Image: Image with bounding boxes and labels drawn.
    """
    draw = ImageDraw.Draw(image)
    font = ImageFont.load_default()

    for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
        # Draw bounding box
        draw.rectangle(box.tolist(), outline="red", width=20)

        # Draw label
        label_text = f"{id2label[label.item()]}: {score:.4f}"
        draw.text((box[0], box[1]), label_text, fill="white", font=font)

    return image