Files
wifi-densepose/references/wifi_densepose_results.csv
rUv f3c77b1750 Add WiFi DensePose implementation and results
- Implemented the WiFi DensePose model in PyTorch, including CSI phase processing, modality translation, and DensePose prediction heads.
- Added a comprehensive training utility for the model, including loss functions and training steps.
- Created a CSV file to document hardware specifications, architecture details, training parameters, performance metrics, and advantages of the model.
2025-06-07 05:23:07 +00:00

4.4 KiB

1CategoryMetricValueUnitDescription
2HardwareWiFi_Transmitters3countNumber of WiFi transmitter antennas
3HardwareWiFi_Receivers3countNumber of WiFi receiver antennas
4HardwareFrequency_Range2.4GHz ± 20MHzfrequencyOperating frequency range
5HardwareSubcarriers30countNumber of subcarrier frequencies
6HardwareSampling_Rate100HzCSI data sampling rate
7HardwareTotal_Cost30USDHardware cost using TP-Link AC1750 routers
8ArchitectureInput_Amplitude_Shape150x3x3tensorCSI amplitude input dimensions
9ArchitectureInput_Phase_Shape150x3x3tensorCSI phase input dimensions
10ArchitectureOutput_Feature_Shape3x720x1280tensorSpatial feature map dimensions
11ArchitectureBody_Parts24countNumber of body parts detected
12ArchitectureKeypoints17countNumber of keypoints tracked (COCO format)
13TrainingLearning_Rate0.001rateInitial learning rate
14TrainingBatch_Size16countTraining batch size
15TrainingTotal_Iterations145000countTotal training iterations
16TrainingLambda_DensePose0.6weightDensePose loss weight
17TrainingLambda_Keypoint0.3weightKeypoint loss weight
18TrainingLambda_Transfer0.1weightTransfer learning loss weight
19PerformanceWiFi_Same_Layout_AP43.5APAP for WiFi_Same_Layout
20PerformanceWiFi_Same_Layout_AP@5087.2APAP@50 for WiFi_Same_Layout
21PerformanceWiFi_Same_Layout_AP@7544.6APAP@75 for WiFi_Same_Layout
22PerformanceWiFi_Same_Layout_AP-m38.1APAP-m for WiFi_Same_Layout
23PerformanceWiFi_Same_Layout_AP-l46.4APAP-l for WiFi_Same_Layout
24PerformanceWiFi_Same_Layout_dpAP_GPS45.3APdpAP_GPS for WiFi_Same_Layout
25PerformanceWiFi_Same_Layout_dpAP_GPS@5079.3APdpAP_GPS@50 for WiFi_Same_Layout
26PerformanceWiFi_Same_Layout_dpAP_GPS@7547.7APdpAP_GPS@75 for WiFi_Same_Layout
27PerformanceWiFi_Same_Layout_dpAP_GPSm43.2APdpAP_GPSm for WiFi_Same_Layout
28PerformanceWiFi_Same_Layout_dpAP_GPSm@5077.4APdpAP_GPSm@50 for WiFi_Same_Layout
29PerformanceWiFi_Same_Layout_dpAP_GPSm@7545.5APdpAP_GPSm@75 for WiFi_Same_Layout
30PerformanceImage_Same_Layout_AP84.7APAP for Image_Same_Layout
31PerformanceImage_Same_Layout_AP@5094.4APAP@50 for Image_Same_Layout
32PerformanceImage_Same_Layout_AP@7577.1APAP@75 for Image_Same_Layout
33PerformanceImage_Same_Layout_AP-m70.3APAP-m for Image_Same_Layout
34PerformanceImage_Same_Layout_AP-l83.8APAP-l for Image_Same_Layout
35PerformanceImage_Same_Layout_dpAP_GPS81.8APdpAP_GPS for Image_Same_Layout
36PerformanceImage_Same_Layout_dpAP_GPS@5093.7APdpAP_GPS@50 for Image_Same_Layout
37PerformanceImage_Same_Layout_dpAP_GPS@7586.2APdpAP_GPS@75 for Image_Same_Layout
38PerformanceImage_Same_Layout_dpAP_GPSm84.0APdpAP_GPSm for Image_Same_Layout
39PerformanceImage_Same_Layout_dpAP_GPSm@5094.9APdpAP_GPSm@50 for Image_Same_Layout
40PerformanceImage_Same_Layout_dpAP_GPSm@7586.8APdpAP_GPSm@75 for Image_Same_Layout
41PerformanceWiFi_Different_Layout_AP27.3APAP for WiFi_Different_Layout
42PerformanceWiFi_Different_Layout_AP@5051.8APAP@50 for WiFi_Different_Layout
43PerformanceWiFi_Different_Layout_AP@7524.2APAP@75 for WiFi_Different_Layout
44PerformanceWiFi_Different_Layout_AP-m22.1APAP-m for WiFi_Different_Layout
45PerformanceWiFi_Different_Layout_AP-l28.6APAP-l for WiFi_Different_Layout
46PerformanceWiFi_Different_Layout_dpAP_GPS25.4APdpAP_GPS for WiFi_Different_Layout
47PerformanceWiFi_Different_Layout_dpAP_GPS@5050.2APdpAP_GPS@50 for WiFi_Different_Layout
48PerformanceWiFi_Different_Layout_dpAP_GPS@7524.7APdpAP_GPS@75 for WiFi_Different_Layout
49PerformanceWiFi_Different_Layout_dpAP_GPSm23.2APdpAP_GPSm for WiFi_Different_Layout
50PerformanceWiFi_Different_Layout_dpAP_GPSm@5047.4APdpAP_GPSm@50 for WiFi_Different_Layout
51PerformanceWiFi_Different_Layout_dpAP_GPSm@7526.5APdpAP_GPSm@75 for WiFi_Different_Layout
52AblationAmplitude_Only_AP39.5APPerformance with amplitude only
53AblationPlus_Phase_AP40.3APPerformance adding phase information
54AblationPlus_Keypoints_AP42.9APPerformance adding keypoint supervision
55AblationFinal_Model_AP43.5APPerformance with transfer learning
56AdvantagesThrough_WallsYesbooleanCan detect through walls and obstacles
57AdvantagesPrivacy_PreservingYesbooleanNo visual recording required
58AdvantagesLighting_IndependentYesbooleanWorks in complete darkness
59AdvantagesLow_CostYesbooleanUses standard WiFi equipment
60AdvantagesReal_TimeYesbooleanMultiple frames per second
61AdvantagesMultiple_PeopleYesbooleanCan track multiple people simultaneously