More contents to be added. Stay tuned!

CSI Tools

Although more and more chip vendors start to support CSI for their own products, most researchers still only have limited access to CSI on commodity WiFi. The below are the currently available tools to obtain CSI on certain chipsets.

  1. Intel 5300 CSI Tool by Daniel Halperin, Wenjun Hu, Anmol Sheth, David Wetherall. This is the first CSI tool that made CSI measurement (on Intel 5300) available to the community has fostered tons of works on WiFi sensing.
  2. Atheros CSI Tool by Yaxiong Xie and Mo Li. Atheros CSI Tool is an open source 802.11n CSI tool working with Qualcomm Atheros WiFi NICs, e.g., AR9580, AR9590, AR9344 and QCA9558.
  3. PicoScenes Wi-Fi Sensing System by Zhiping Jiang's group. According to its homepage, PicoScenes supports QCA9300 and IWL5300.

Public Datasets

Different from CV and NLP areas that have many open datasets, there are few in the community of wireless sensing. Luckily, more and more are showing up!
The following list is to be updated...

Year Project Signal Features Descriptions Dataset
2020 mmGait
(AAAI 2020)
mmWave RSSI, DFS · App.: gait recognition
· Volunteers: 95
· Volume: 30 hours of data
Download
2019 WiAR
(IEEE Access)
WiFi RSSI, CSI · App.: activity recognition
· Volunteers: 10
· Volume: 30 times every volunteer
Download
2019 WiDar 3.0
(MobiSys 2019)
WiFi CSI, DFS, BVP · App.: gesture recognition
· Scenarios: classroom, office and corridor
· Volume:258K instances in 8,620 minutes
Download
2018 WiDar 1.0/2.0
(MobiHoc 2017)
(MobiSys 2018)
WiFi CSI, AoA, DFS · App.: passive localization and tracking
· Scenarios: classroom, office and corridor
· Volume: 80 traces
Download
2018 SignFi
(UbiComp 2018)
WiFi CSI · App.: gesture recognition
· Scenarios: lab and home environment
· Volume: About 6GB 276 sign words
Download
2017 (N/A)
(IEEE Commun. Mag.)
WiFi CSI · App.: activity recognition
· Volunteers: 6
· Volume: About 4GB data of 6 activities
Download

* This table is adapted and extended based on the list compiled by the list RF-based Activity Recognition Datasets.