overwatch-sayak

Overwatch Sayak

Overview

The powers of this library overwatch-sayak package can be used for object detection in images or video, loading datasets, detection tracking, counting, detecting, Slicing Aided Hyper Inferencing for small object detection, etc.

Installation

you need to install the overwatch-sayak package. You can do this using pip:

pip install overwatch-sayak

For importing the library for your code, use the import command with:

import overwatch_sayak

Quickstart

Models

overwatch-sayak was designed to be model-friendly. Just plug in any classification, detection, or segmentation model. We have created connectors for the most popular libraries like Ultralytics, Transformers, or MMDetection for your convenience.

Inference

import cv2
import overwatch_sayak as ov
from inference import get_model

image = cv2.imread("path/to/your/image.jpg")
model = get_model("yolov8s-640")
result = model.infer(image)[0]
detections = ov.Detections.from_inference(result)

len(detections)
# Output: Number of detections

Annotators

overwatch-sayak offers a wide range of highly customizable annotators, allowing you to compose the perfect visualization for your use case.

import cv2
import overwatch_sayak as ov

image = cv2.imread("path/to/your/image.jpg")
detections = ov.Detections(...)

bounding_box_annotator = ov.BoundingBoxAnnotator()
annotated_frame = bounding_box_annotator.annotate(
    scene=image.copy(),
    detections=detections
)

# Display or save the annotated image
cv2.imshow("Annotated Image", annotated_frame)
cv2.waitKey(0)
cv2.destroyAllWindows()
# Or save the image
cv2.imwrite("annotated_image.jpg", annotated_frame)

Slicing Inference for Small Object Detection

Slicing the image into smaller pieces can improve object detection accuracy for detecting small objects. The following example demonstrates how to use the InferenceSlicer for this purpose.

import overwatch_sayak as ov
from inference import get_model
import cv2
import numpy as np

# Load the image
image = cv2.imread("/path/to/image.jpg")

# Load the model
model = get_model("yolov8s-640")

# Define the callback function for slicing inference
def slicer_callback(slice: np.ndarray) -> ov.Detections:
    result = model.infer(slice)[0]
    detections = ov.Detections.from_inference(result)
    return detections

# Create the slicer
slicer = ov.InferenceSlicer(
    callback=slicer_callback,
    slice_wh=(512, 512),
    overlap_ratio_wh=(0.4, 0.4),
    overlap_filter_strategy=ov.OverlapFilter.NONE
)

# Run the slicer on the image
detections = slicer(image)

# Annotate the image
annotated_frame = ov.BoundingBoxAnnotator().annotate(
    scene=image.copy(),
    detections=detections
)

# Display or save the annotated image
cv2.imshow("Annotated Image", annotated_frame)
cv2.waitKey(0)
cv2.destroyAllWindows()
# Or save the image
cv2.imwrite("annotated_image.jpg", annotated_frame)

Datasets

overwatch-sayak offers a suite of utilities that enable you to load, split, merge, and save datasets in various supported formats..

import overwatch_sayak as ov

dataset = ov.DetectionDataset.from_yolo(
    images_directory_path="path/to/images",
    annotations_directory_path="path/to/annotations",
    data_yaml_path="path/to/data.yaml"
)

dataset.classes
# Output: ['dog', 'person']

len(dataset)
# Output: Number of images in the dataset

Split

train_dataset, test_dataset = dataset.split(split_ratio=0.7)
test_dataset, valid_dataset = test_dataset.split(split_ratio=0.5)

len(train_dataset), len(test_dataset), len(valid_dataset)
# Output: (Number of training images, Number of test images, Number of validation images)

Merge

ds_1 = ov.DetectionDataset(...)
len(ds_1)
# Output: Number of images in ds_1
ds_1.classes
# Output: ['dog', 'person']

ds_2 = ov.DetectionDataset(...)
len(ds_2)
# Output: Number of images in ds_2
ds_2.classes
# Output: ['cat']

ds_merged = ov.DetectionDataset.merge([ds_1, ds_2])
len(ds_merged)
# Output: Number of images in the merged dataset
ds_merged.classes
# Output: ['cat', 'dog', 'person']

Save

dataset.as_yolo(
    images_directory_path="path/to/save/images",
    annotations_directory_path="path/to/save/annotations",
    data_yaml_path="path/to/save/data.yaml"
)

dataset.as_pascal_voc(
    images_directory_path="path/to/save/images",
    annotations_directory_path="path/to/save/annotations"
)

dataset.as_coco(
    images_directory_path="path/to/save/images",
    annotations_path="path/to/save/annotations"
)

Convert

ov.DetectionDataset.from_yolo(
    images_directory_path="path/to/load/images",
    annotations_directory_path="path/to/load/annotations",
    data_yaml_path="path/to/load/data.yaml"
).as_pascal_voc(
    images_directory_path="path/to/save/images",
    annotations_directory_path="path/to/save/annotations"
)

Contributing

If you want to contribute to this project, feel free to open an issue or submit a pull request on GitHub.

License

This project is licensed under the MIT License.