Data for Image Recognition Tasks: An Efficient Tool for Fine-Grained Annotations

Filax, Marco; Gonschorek, Tim; Ortmeier, Frank

doi:10.5220/0007688709000907

Cited by 8 publications

(6 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In our opinion, fine-grained datasets cover a very challenging acquisition problem. Thus, we chose to refine a semiautomatic generated large-scale dataset taken from [7] to illustrate the efficiency of the proposed approach.…”

Section: A Retail Recognition Datasetmentioning

confidence: 99%

“…We find it challenging to determine the fine-grained visual differences of retail products because products typically share significant visual aspects and the enormous reference concept space. This makes the Magdeburg Groceries Dataset [7] the perfect basis for our experiments.…”

Section: A Retail Recognition Datasetmentioning

confidence: 99%

“…It allows the labeler to monitor the labeling progress itself and the distribution of tagged candidate streams. Further, it displays a comparison to the original Magdeburg Grocery dataset [7]. We identified 1188 visual concepts, i.e., retail products.…”

Section: Manual Annotationmentioning

confidence: 99%

“…Using additional meta-knowledge, i.e., the depth stream of a video stream, [21] proposed to track the segmented object in 3D space and reproject bounding boxes onto the image planes. [7] used meta-knowledge of the environment to propose a semi-automatic annotation tool tailored for retail environments. The authors used a SLAM approach to annotate objects in 3D space and reproject their annotation onto consecutive video frames.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Semi-automatic Acquisition of Datasets for Retail Recognition

Filax¹,

Gonschorek²

2022

CSRN

Self Cite

View full text Add to dashboard Cite

The acquisition of datasets is typically a laborious task. It is challenging, especially if the required annotations in every image in the dataset are vast. It is even more challenging if the inter-class variance, the visual difference between two distinct classes, is low. Retail product recognition constitutes an example of both issues. Products are densely packed on shelves, resulting in many objects within an image. Products share visual similarities, which makes them hard to distinguish. In this work, we propose Annotron, a tool tackling the acquisition problem in this domain. Exploiting dataset struc- tures, such as being organized in consecutive frames, we detect real-world objects through pre-trained detectors and reproject detections to generate candidate traces over time. Further, we aid labelers by computing poten- tial matches of real-world objects and reference images based on their visual similarity: We cluster consecutive detections based on a large set of reference images using embeddings acquired from pre-trained networks. Using the proposed tool reduces manual efforts drastically by diminishing the time spent on repetitive, error-prone tasks. We evaluate Annotron in the retail recognition domain. The domain is commonly considered fine-grained, which means that instance-level annotations are costly due to the described problems. We refine the given dataset, surpass the number of previously found stock-keeping units, and label over 446.500 individual bounding boxes.

show abstract

Section: A Retail Recognition Datasetmentioning

confidence: 99%

Section: A Retail Recognition Datasetmentioning

confidence: 99%

Section: Manual Annotationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Semi-automatic Acquisition of Datasets for Retail Recognition

Filax¹,

Gonschorek²

2022

CSRN

Self Cite

View full text Add to dashboard Cite

show abstract

“…Over the next decade, 2D grocery datasets [12,14] mainly improved the RGB images' resolution. The improvement in computing hardware during the past 15 years has paved the way to larger datasets [16,17,19,21,23,25,29,[31][32][33], gradually increasing the number of categories and images per category. The Supermarket Produce [16] dataset features varied product view angles but uses a white canvas background, unlike real grocery store settings.…”

Section: D Grocery Datasetsmentioning

confidence: 99%

Dynamic Local Geometry Capture in 3D Point Cloud Classification

Sheshappanavar

Kambhamettu

2021

2021 IEEE 4th International Conference on Multimedia Information Processing and Retrieval (MIPR)

View full text Add to dashboard Cite

Fine-grained grocery object recognition is an important computer vision problem with broad applications in automatic checkout, in-store robotic navigation, and assistive technologies for the visually impaired. Existing datasets on groceries are mainly 2D images. Models trained on these datasets are limited to learning features from the regular 2D grids. While portable 3D sensors such as Kinect were commonly available for mobile phones, sensors such as LiDAR and TrueDepth, have recently been integrated into mobile phones. Despite the availability of mobile 3D sensors, there are currently no dedicated real-world largescale benchmark 3D datasets for grocery. In addition, existing 3D datasets lack fine-grained grocery categories and have limited training samples. Furthermore, collecting data by going around the object versus the traditional photo capture makes data collection cumbersome. Thus, we introduce a large-scale grocery dataset called 3DGro-cery100. It constitutes 100 classes, with a total of 87,898 3D point clouds created from 10,755 RGB-D single-view images. We benchmark our dataset on six recent stateof-the-art 3D point cloud classification models. Additionally, we also benchmark the dataset on few-shot and continual learning point cloud classification tasks. Project Page: https://bigdatavision.org/3DGrocery100/.

show abstract