VisionCounter is a powerful Python library that provides a collection of deep learning models for accurately counting objects in input images. Whether you need to count people in a crowded street, cars in a parking lot, or any other objects in a scene, VisionCounter has got you covered.
-
State-of-the-Art Models: VisionCounter comes with pre-trained state-of-the-art deep learning models, trained on large-scale datasets, to deliver high accuracy in counting various objects.
-
Easy-to-Use API: We believe in simplicity and ease of use. VisionCounter provides a straightforward API, allowing you to seamlessly integrate object counting capabilities into your Python projects.
-
Customizable: While the default pre-trained models work well in many scenarios, we understand that every project is unique. VisionCounter allows you to fine-tune models or even train your custom models on your specific datasets.
-
Fast and Efficient: Our models are designed for speed and efficiency, making them suitable for real-time applications or processing large batches of images.
-
Wide Range of Applications: VisionCounter is versatile and can be used in various domains, including surveillance, traffic monitoring, inventory management, and more.
Install the library using pip
pip install vision_counter
Usage
from sys import argv
from vision_counter import CounTR
counter = CounTR()
bboxes = [[[156, 112], [173, 129]]] # x1, y1, x2, y2
image_path = argv[1]
object_count = counter.count(image_path, bboxes)
print("Estimated object count:", object_count)We welcome contributions from the community! Whether it's bug fixes, feature enhancements, or new models, feel free to submit pull requests.
@article{liu2022countr,
author = {Chang, Liu and Yujie, Zhong and Andrew, Zisserman and Weidi, Xie},
title = {CounTR: Transformer-based Generalised Visual Counting},
journal = {arXiv:2208.13721},
year = {2022}
}