usgs_extract

Object Detection Pipeline

Functions

MaxResize

A custom image transformation class that resizes an image such that its largest dimension does not exceed a specified maximum size, preserving the aspect ratio.

Attributes:

• max_size (int): The maximum allowed size for the largest dimension of the image.

Methods:

• __init__(self, max_size=800): Initializes the transformation with a given maximum size.
• __call__(self, image): Resizes the input image to ensure its largest dimension does not exceed max_size.

Usage:

resize_transform = MaxResize(800)
resized_image = resize_transform(image)

detection_transform

A composed transformation that applies MaxResize, converts the image to a tensor, and normalizes it using predefined mean and standard deviation values.

Composition:

• MaxResize(800)
• transforms.ToTensor()
• transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])

Usage:

detection_transform = transforms.Compose([
    MaxResize(800),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])

box_cxcywh_to_xyxy(x)

Converts bounding boxes from center x, center y, width, height format to x_min, y_min, x_max, y_max format.

Parameters:

• x (torch.Tensor): Tensor containing bounding boxes in (cx, cy, w, h) format.

Returns:

• torch.Tensor: Tensor containing bounding boxes in (x_min, y_min, x_max, y_max) format.

Usage:

bbox_xyxy = box_cxcywh_to_xyxy(bbox_cxcywh)

rescale_bboxes(out_bbox, size)

Rescales bounding boxes to the original image size.

Parameters:

• out_bbox (torch.Tensor): Tensor containing bounding boxes in (cx, cy, w, h) format.
• size (tuple): The original size of the image as (width, height).

Returns:

• torch.Tensor: Tensor containing rescaled bounding boxes in (x_min, y_min, x_max, y_max) format.

Usage:

rescaled_bboxes = rescale_bboxes(bboxes, image.size)

outputs_to_objects(outputs, img_size, id2label)

Processes the model outputs to extract detected objects, their labels, scores, and bounding boxes.

Parameters:

• outputs (dict): The output from the object detection model.
• img_size (tuple): The original size of the image as (width, height).
• id2label (dict): Dictionary mapping label ids to label names.

Returns:

• list: A list of dictionaries, each containing the label, score, and bounding box of a detected object.

Usage:

objects = outputs_to_objects(outputs, image.size, id2label)

runModel(files, directory_path)

Processes a list of image files to detect objects and saves the results in a CSV file.

Parameters:

• files (list): A list of file paths to the images.
• directory_path (str): The directory path to save the results CSV file.

Returns:

• pd.DataFrame: A DataFrame containing the file paths and the detected objects with their scores.

Usage:

files = ["path/to/image1.jpg", "path/to/image2.jpg"]
results = runModel(files, "path/to/save/results")
print(results)

---

This README provides a detailed guide to understanding and utilizing the functions in the script for object detection and processing.