Newer
Older
# The large-scale retail scenario object detection and counting task (Locount)
[Rethinking Object Detection in Retail Stores](http://arxiv.org/abs/2003.08230).
## Problem definition
## Locount dataset
## Evaluation protocol
To fairly compare algorithms on the *Locount* task, we design a new evaluation protocol, which penalizes algorithms for missing object instances, for duplicate detections of one instance, for false positive detections, and for false counting numbers of detections. Inspired by MS COCO protocol, we design new metrics *AP^{lc}*, *AP_{0.5}^{lc}*, *AP_{0.75}^{lc}*, and *AR^{lc}_{max}=150}* to evaluate the performance of methods, which takes both the localization and counting accuracies into account. Specifically, a correct detection should satisfied two criteria, (1) the localization intersection over union, $\text{IoU} = \frac{\widehat{B} \cap B^\ast}{ \widehat{B} \cup B^\ast}$, between the predicted bounding box $\widehat{B}$ and the ground-truth bounding box $B^\ast$ is larger than the threshold $\theta_{\text{l}}$, \ie, $\text{IoU} \geq \theta_{\text{l}}$; and (2) the counting accuracy, $\text{AC} = \max\big(0, 1-\frac{|\widehat{C} - C^\ast|}{C^\ast} \big)$, between the predicted instance number enclosed in the predicted bounding box $\widehat{C}$ and the ground-truth instance number $C^\ast$ is larger than the threshold $\theta_{\text{c}}$, \ie, $\text{AC} \geq \theta_{\text{c}}$. After that, AP$^{\it lc}$ is computed by averaging over all $10$ IoU thresholds, \ie, $\theta_{\text{l}} \in [0.50, 0.95]$ with the uniform step size $0.05$, and $10$ AC thresholds, \ie, $\theta_{\text{c}} \in [0.50, 0.95]$ with the uniform step size $0.05$, of all categories, which is used as the primary metric for ranking algorithms. Note that, to indicate the localization accuracy, we also report the results of evaluated methods using the MS COCO protocol for reference.
## Baseline method
The convention standard for object detection uses a bounding box to represent each individual object instance. However, it is not practical in the industry-relevant applications in the context of warehouses due to severe occlusions among groups of instances of the same categories. In this paper, we propose a new task, ie, simultaneously object localization and counting, abbreviated as *Locount*, which requires algorithms to localize groups of objects of interest with the number of instances. However, there does not exist a dataset or benchmark designed for such a task. To this end, we collect a large-scale object localization and counting dataset with rich annotations in retail stores, which consists of 50,394 images with more than 1.9 million object instances in 140 categories. Together with this dataset, we provide a new evaluation protocol and divide the training and testing subsets to fairly evaluate the performance of algorithms for Locount, developing a new benchmark for the Locount task. Moreover, we present a cascaded localization and counting network as a strong baseline, which gradually classifies and regresses the bounding boxes of objects with the predicted numbers of instances enclosed in the bounding boxes, trained in an end-to-end manner. Extensive experiments are conducted on the proposed dataset to demonstrate its significance and the analysis discussions on failure cases are provided to indicate future directions.
## Citation
If you find this dataset useful for your research, please cite
```
@inproceedings{Cai2020Locount,
title={Rethinking Object Detection in Retail Stores},
author={Yuanqiang Cai and Longyin Wen and Libo Zhang and Dawei Du and Weiqiang Wang},
booktitle={arXiv preprint arXiv:2003.08230},
year={2020}
}
```
The *Locount* dataset is formed by 50,394 JPEG images with the resolution of 1920 × 1080 pixels.
The *training* subset includes 34,022 images.
The *testing* subset includes 16,372 images.
## Evaluation
To fairly compare algorithms on the *Locount* task, we design a new evaluation protocol, which penalizes algorithms for missing object instances, for duplicate detections of one instance, for false positive detections, and for false counting numbers of detections. Please check our paper for details.
Suggestions and opinions of this dataset are welcome. Please contact the authors by sending email to libo@iscas.ac.cn.