Supervised learning for Out-of-Stock detection in panoramas of retail shelves

Abstract

Improving inventory management is essential to retailer profitability. This paper proposes a supervised learning approach for Out-of-Stock (OOS) detection by Texture, Color and Geometry features in high-resolution panoramic images of grocery retail shelves. Cascade classifiers are used to detect labels that can potentially be used to confirm the presence of the OOS cases. The image acquisition setup includes a camera cart that shoots from multi-viewpoints aiming a parallel motion to the shelf. The correction of perspective distortion is applied to handle the different camera translation motions while stitching together images with a high-level of similarity. From the generated panoramas, the proposed OOS detection is followed by classification with Support Vector Machines. The experimental tests were performed throughout the retail environment with real data obtained from supermarket shelves containing labels near the visible ruptures. Results show a detection accuracy of 84.5% for OOS and a sensitivity of 86.6% for label detection

IoT Sensing Platform as a Driver for Digital Farming in Rural Africa

“…The soil particles distribution is used for soil classification and thus serves as the estimator for water holding capacity and drainage, aeration, susceptibility to erosion and cation exchange capacity, pH buffering capacity. Alternatively, using the library Texture-Color-Geometry Feature Extraction (TCGFE) [ 36 , 37 , 38 ] developed by Fraunhofer Portugal AICOS, a total of 152 features can be extracted for each region of interest in the image, which can later be used for machine learning purposes. Alternatively, texture can also be indirectly measured by fractal dimension [ 39 ].…”

Towards Intelligent Retail: Automated on-Shelf Availability Estimation Using a Depth Camera

“…More recently, computer vision systems have been proposed by many as a promising solution for smart retail applications including detection of misplaced products [11], verification of planogram compliance [6], and stock assessment [12]. Cameras for data gathering can be installed at fixed locations in the store [6], or integrated into smartphones [13], manually-driven carts [12] or on-board mobile robotic platforms [14]. Image processing techniques mainly using state-of-the-art feature-based or template-based matching algorithms have been adopted to look for anomalies and trigger alarms for store managers.…”

“…In order to cope with the high variability of store environments, the use of machine learning techniques has been proposed, to ensure higher robustness and accuracy, although at the cost of labor-intensive manual image annotation for training. An example can be found in [12], where a supervised learning approach based on Support Vector Machines (SVM) is developed for OOS detection in panoramic images of retail shelves. The camera image setup includes a camera cart that is moved parallel to the shelf to acquire images from multiple viewpoints.…”

Robust Shelf Monitoring Using Supervised Learning for Improving On-Shelf Availability in Retail Stores

“…Methods [3,4] have recently been proposed to monitor the shelves by using the images for retail stores. Kejriwal et al [3] proposed a method to directly obtain product count from images using mobile robots for stock monitoring and assessment in retail stores.…”

“…Hence, it requires a large database of high-quality product templates, which is labor-intensive to maintain since products frequently change due to new and seasonal products arriving. Rosado et al [4] proposed a method to detect out-of-stock regions in grocery retail shelves without using a large database. This method detects out-of-stock regions using points of interest (keypoints) detected by FAST (Features from Accelerated Segment Test) [6].…”