REFORM: Recognizing F-formations for Social Robots

Hedayati, Hooman; Muehlbradt, Annika; Szafir, Daniel; Andrist, Sean

doi:10.1109/iros45743.2020.9340708

Cited by 12 publications

(4 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As we can see, our simulation dataset can improve the F1 score on the SALSA dataset. Our best achievement on the SALSA dataset is nearly 72.61 F1@2/3, which is comparable to 81.2 from [9].…”

Section: Methodsmentioning

confidence: 62%

“…A deep affinity network predicts the graph's affinity matrix, and the Dominant Sets algorithm is then used to find F-formation clusters. Meanwhile, Hedayati et al [9] proposed a similar method, introducing Distance and Effort Angle as new input features. The problem formulation also involves building a graph to represent the connections between the agents, and the affinity matrix is predicted using Weighted KNN, Bagged Trees and Logistic Regression.…”

Section: Methodsmentioning

confidence: 99%

“…Available datasets such as [4] do not contain information on bounding boxes, and there is a lack of variety in condition, camera setup, people and quantity. Furthermore, various approaches have been made toward the F-formation detection task, ranging from traditional [6] [7] to data-driven methods [8] [9]. The major drawbacks of these works are (1) the need to use additional devices (sensors, depth camera,...) to capture intermediary information like body pose or head orientation or (2) the reliance on heuristically determined threshold values.…”

Section: Introductionmentioning

confidence: 99%

“…Most methods rely on hard-coded parameters whose values are derived by experiments or conventional methods from small-scale datasets, which may not be generalizable. For example, [8], [9] are experimented on SALSA dataset [4], [5], etc. These datasets are fixed at the number of people in the frame and camera setup as well as the environment.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

F2SD: a dataset for end-to-end group detection algorithms

Hoang

Nguyen

Cao

et al. 2023

Fifteenth International Conference on Machine Vision (ICMV 2022)

View full text Add to dashboard Cite

The lack of large-scale datasets has been impeding the advance of deep learning approaches to the problem of F-formation detection. Moreover, most research on this problem rely on input sensor signals of object location and orientation rather than image signals. To address this, we develop a new, large-scale dataset of simulated images for F-formation detection, called F-formation Simulation Dataset (F2SD). F2SD contains nearly 60,000 images simulated from GTA-5, with bounding boxes and orientation information on images, making it useful for a wide variety of modelling approaches. It is also closer to practical scenarios, where three-dimensional location and orientation information are costly to record. It is challenging to construct such a large-scale simulated dataset while keeping it realistic. Furthermore, the available research utilizes conventional methods to detect groups. They do not detect groups directly from the image. In this work, we propose (1) a large-scale simulation dataset F2SD and a pipeline for F-formation simulation, (2) a first-ever end-to-end baseline model for the task, and experiments on our simulation dataset.

show abstract

Section: Methodsmentioning

confidence: 62%

Section: Methodsmentioning

confidence: 99%