Multi-View Radar Semantic Segmentation

Ouaknine, Arthur; Newson, Alasdair; Pérez, Patrick; Tupin, Florence; Rebut, Julien

doi:10.1109/iccv48922.2021.01538

Cited by 52 publications

(41 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is due to fact that IoT device 5 has camera and radar sensors and is using both Mask R-CNN and TMVA-NET algorithms for semantic extraction which have high quality of extracting significant semantic information. The remaining element from group 1 is IoT device 2 uses RAMP-CNN for radar semantic extraction, which has the lowest performance compared to other algorithms [9].…”

Section: B Resultsmentioning

confidence: 99%

“…Specifically, we propose the transmission of only semantic information to the VSP instead of raw data. Since the semantic extraction from raw data depends heavily on the data type, we focus in this work on the case of camera and radar sensors for AV systems [8], [9]. This is motivated by the fact that camera and radar sensors produce data with high volume and consume a large amount of bandwidth for transmission.…”

Section: B Contributionmentioning

confidence: 99%

“…For devices with camera sensors only, Mask R-CNN algorithm for semantic segmentation and bounding boxes is used [15]. For devices with radar sensors only, RAMP-CNN (for RA view only), FCN-8s (for either RA or RD views) and, TMVA-NET (for RA and RD views) are used [9], [14], [16]. We adopt simulation settings similar to [9].…”

Section: A Dataset and Experimentation's Settingsmentioning

confidence: 99%

“…For devices with radar sensors only, RAMP-CNN (for RA view only), FCN-8s (for either RA or RD views) and, TMVA-NET (for RA and RD views) are used [9], [14], [16]. We adopt simulation settings similar to [9]. Each IoT device is considered to use one of the aforementioned algorithms based on a uniform distribution.…”

Section: A Dataset and Experimentation's Settingsmentioning

confidence: 99%

See 3 more Smart Citations

Semantic Information Market For The Metaverse: An Auction Based Approach

Lotfi¹,

Niyato²,

Sun³

et al. 2022

Preprint

View full text Add to dashboard Cite

In this paper, we address the networking and communications problems of creating a digital copy in the Metaverse digital twin. Specifically, a virtual service provider (VSP) which is responsible for creating and rendering the Metaverse, is required to use the data collected by IoT devices to create the virtual copy of the physical world. However, due to the huge volume of the collected data by IoT devices (e.g., images and videos) and the limited bandwidth, the VSP might become unable to retrieve all the required data from the physical world. Furthermore, the Metaverse needs fast replication (e.g., rendering) of the digital copy adding more restrictions on the data transmission delay. To solve the aforementioned challenges, we propose to equip the IoT devices with semantic information extraction algorithms to minimize the size of the transmitted data over the wireless channels. Since many IoT devices will be interested to sell their semantic information to the VSP, we propose a truthful reverse auction mechanism that helps the VSP select only IoT devices that can improve the quality of its virtual copy of objects through the semantic information. We conduct extensive simulations on a dataset that contains synchronized camera and radar images, and show that our novel design enables a fast replication of the digital copy with high accuracy.

show abstract

Section: B Resultsmentioning

confidence: 99%

Section: B Contributionmentioning

confidence: 99%

Section: A Dataset and Experimentation's Settingsmentioning

confidence: 99%

Section: A Dataset and Experimentation's Settingsmentioning

confidence: 99%

See 2 more Smart Citations

Semantic Information Market For The Metaverse: An Auction Based Approach

Lotfi¹,

Niyato²,

Sun³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…The RA view has been a research topic for multi-class [16] and free space [29] segmentation. The entire RAD tensor is considered for multi-view segmentation in [30]. Radar point cloud segmentation has also been explored to estimate bird-eye-view occupancy grids, either for LD [22], [39] or HD [34], [33], [37] radars.…”

Section: Radar Backgroundmentioning

confidence: 99%

Raw High-Definition Radar for Multi-Task Learning

Rebut¹,

Ouaknine²,

Malik³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Fig. 1: Overview of our RADIal dataset. RADIal includes a set of 3 sensors (camera, laser scanner, high-definition radar) and comes with GPS and vehicle's CAN traces; 25K synchronized samples are recorded in raw format. (a) Camera image with projected laser point cloud in red and radar point cloud in indigo, vehicle annotation in orange and free driving space annotation in green; (b) Radar power spectrum with bounding box annotations; (c) Free driving space annotation in bird-eye view, with annotated vehicle bounding boxes in orange, radar point cloud in indigo and laser point cloud in red; (d) Range-Azimuth map in Cartesian coordinates overlayed with radar point cloud and laser point cloud; (e) GPS trace in red and odometry trajectory reconstruction in green.

show abstract