Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for...
Transcript of Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for...
![Page 1: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/1.jpg)
Video-Based Activity Forecasting
for Construction Safety Monitoring Use Cases
Speaker: Shuai TangUniversity of Illinois at Urbana-Champaign
GTC, Santa Jose, 2019
Contributors: Mani Golparvar Fard (University of Illinois)
Milind Naphade (Nvidia), Murali Gopalakrishna (Nvidia), Amit Goel (Nvidia)
![Page 2: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/2.jpg)
Worker Dies From Falling 50 Feet
2References: California FACE Report #07CA009
![Page 3: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/3.jpg)
33
14 Worker Deaths Every Day In The US
20.7% of all worker deaths were in construction
OSHA estimates that eliminating top 4 hazards in construction
save 581 workersā lives
Falls: 381 deaths (39.2%)
Struck By Object: 80 deaths (8.2%)
Electrocution: 71 deaths (7.3%)
Caught-In/Between: 50 deaths (5.1%)
![Page 4: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/4.jpg)
44
![Page 5: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/5.jpg)
'Carelessā Operator Crushes Worker With Backhoe
5References: https://www.cdc.gov/niosh/topics /highwayworkzones/bad/pdfs/catre port2.pdf
![Page 6: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/6.jpg)
Non-fatal Injuries In Construction
āŖ Safety incidentsāŖ 971 fatal casesāŖ 79,810 non-fatal cases involving days away from work
āŖ $1.3 trillion construction expenditure each year
āŖ Financial impact of safetyāŖ Around $4 million cost per fatal case,āŖ Over $42,000 average cost per non-fatal case.
6
![Page 7: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/7.jpg)
Motivation
7
Frequency
Safety inspections are taken typically weekly.
Accuracy
50% hazards not recognized by workers
Proactiveness
Safety measurements are often retrospective
Image sources: Google Image
![Page 8: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/8.jpg)
Overreaching Goal:Visual-based activity forecasting towards
predictive safety monitoring
![Page 9: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/9.jpg)
Opportunity - Growth In Visual Data
9
200-1,000 pictures per day ~1,000 pictures per day Time-Lapse pictures every 5-15min
~2,000 images per week 1-5 scans/month1-10 videos per day
![Page 10: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/10.jpg)
10Video sources: RAAMAC Lab
![Page 11: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/11.jpg)
11Video sources: Jun Yang, RAAMAC Lab
![Page 12: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/12.jpg)
Video sources: RAAMAC Lab
![Page 13: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/13.jpg)
Documentation, Intervention, Near-Miss Reporting
13
Near-miss ReportingRight-time Intervention
Image sources: Left: http://www.energysafetycanada.com/files/safety-alerts/Safety%20Alert%20-%2010.2018%20-%20Final.pdf Right: http://www.energysafetycanada.com/files/safety-alerts/Safety%20Alert-13%202016.pdf
![Page 14: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/14.jpg)
1414
Big Picture - Computer Vision & Jobsite Cameras
Detect, Track, Model Worker Activities
Understand Work Context
Predict Next Sequence of Activities
![Page 15: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/15.jpg)
Social LSTM (Alahi et al. 2016)
15
Vanilla LSTM (Graves, 2013)T= 1 T= 2
x1y1x2y2
15Slide: Alexandre Alahi
Activity Forecasting ā Computer Vision
![Page 16: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/16.jpg)
13
Social LSTMT= 1 T= 2
Details on social pooling for person 2 (in white)
Top view16
Activity Forecasting ā Computer Vision
Social LSTM (Alahi et al. 2016)
Slide: Alexandre Alahi
![Page 17: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/17.jpg)
13
Social LSTMT= 1 T= 2
Details on social pooling for person 2 (in white)
17
Activity Forecasting ā Computer Vision
Social LSTM (Alahi et al. 2016)
Occupancy mapSlide: Alexandre Alahi
Top view
![Page 18: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/18.jpg)
13
Social LSTMT= 1 T= 2
Details on social pooling for person 2 (in white)
18
Activity Forecasting ā Computer Vision
Social LSTM (Alahi et al. 2016)
Slide: Alexandre Alahi
h1
h3
H2
Social tensorOccupancy mapTop view
![Page 19: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/19.jpg)
Social LSTM learned to turn around a group
1
16
19Slide: Alexandre Alahi
Activity Forecasting ā Computer Vision
Social LSTM (Alahi et al. 2016)ā Black line is the ground truth trajectoryā Gray line is the pastā Heatmap is the predicted distribution
![Page 20: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/20.jpg)
Social LSTM (Alahi et al. 2016)ā Black line is the ground truth trajectoryā Gray line is the pastā Heatmap is the predicted distribution
Social LSTM learned to turn around a group
1
16
20Slide: Alexandre Alahi
Activity Forecasting ā Computer Vision
![Page 21: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/21.jpg)
From Crowd Scenes To Construction Sites
21
Crowd scenes from UCY and ETH dataset Example construction sites, Google Image
![Page 22: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/22.jpg)
22
Construction sites often change drastically
D1 D5 D10
D14 D19 D21
From Crowd Scenes To Construction Sites
![Page 23: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/23.jpg)
2323
Approach ā data-driven, context rich,
and sequence-to-sequence models
![Page 24: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/24.jpg)
Model Architecture (Social LSTM)
For i āth trajectory at time t ā¦ predict i ās location at t+1
(š„š” , š¦š”) Embed Layer
Social
Feature at t
LSTM Decoder
Mixture Density Network(MDN)
MDN output at t+1
Embed Layer
(š„š”+1, š¦š”+1)
Concatenation
Tensors
Model parameters
24
jāth bivariate Gaussian:
[ šš”+1š
, šš”+1š
, šš”+1š
, šš”+1š
]
![Page 25: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/25.jpg)
Model Architecture (Social LSTM)
(š„š” , š¦š”) Embed Layer
Social
Feature at t
LSTM Decoder
Mixture Density Network(MDN)
MDN output at t+1
Embed Layer
(š„š”+1, š¦š”+1)
Concatenation
Tensors
Model parameters
25
Embed Layer
Social Feature at
t+1
LSTM
DecoderMDN
MDN output at
t+2
Embed Layer
(š„š”+2, š¦š”+2)
.
.
.
.
.
.
(š„š”+š, š¦š”+š)
For i āth trajectory at time t ā¦ predict i ās location at t+1
![Page 26: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/26.jpg)
Model Architecture (Ours)
26
For i āth trajectory at time t ā¦ predict i ās location at {t+s1 , t+s2 , ā¦ , t+sk }
.
.
.
.
.
(š„š” ,š¦š”) LSTM Encoder
OccuMap at t
LSTM Decoder MDN MDN output at t+s1
Trajectory feature at t
MDN output at t+sk
MDN output at t+s2Object Class of i
(š„š”+š1 ,š¦š”+š1)
(š„š”+š2 ,š¦š”+š2)
(š„š”+šš, š¦š”+šš)
Concatenation
Tensors
Model parameters
![Page 27: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/27.jpg)
Model Architecture (Ours) - Occupancy Map
27
![Page 28: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/28.jpg)
Color Code and Movement
Blue South West to North EastLime: North East to South WestRed: East to West
Yellow: North to South
Length: Average length of all
trajectories belonging to the cluster
Thickness: Cluster size (number of
Trajectories in the cluster)
Trajectory Features From Common Trajectories
28
Model Architecture (Ours)
![Page 29: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/29.jpg)
29
Model Architecture (Ours)
Iteratively Use Predicted Locations As Inputs Lead to Large Deviations
![Page 30: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/30.jpg)
Model Architecture (Ours)
30
Concatenation
Tensors
Model parameters[ MDN parameters at t+s]
The jāth Gaussian Parameters:
[ šš”+š š , šš”+š
š , šš”+š š , šš”+š
š ]
Negative Log-likelihood (NLL) over all Gaussians of all traj.
Training time
Iš”+š = argmaxš
šš”+š š
Inference time
(š„š”+š ,š¦š”+š ) = šš”+š š¼š”+š
.
.
.
.
.
(š„š” ,š¦š”) LSTM Encoder
OccuMap at t
LSTM Decoder MDN MDN output at t+s1
Trajectory feature at t
MDN output at t+sk
MDN output at t+s2Object Class of i
(š„š”+š1 ,š¦š”+š1)
(š„š”+š2 ,š¦š”+š2)
(š„š”+šš, š¦š”+šš)
For i āth trajectory at time t ā¦ predict i ās location at {t+s1 , t+s2 , ā¦ , t+sk }
![Page 31: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/31.jpg)
31
1
Image courtesy of Berni de Nina
Case Study At Nvidia Voyager Site
270 m (887 ft.) by 34 m (110 ft.)
![Page 32: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/32.jpg)
Experiment Setup
āŖ Voyager dataset:
ā¢ 1,464 mins (24.4 hrs) of 1080p videos
ā¢ Trainval set (from 76 clips): person 1630, vehicle 1752
ā¢ Test set (from 29 clips): person 143, vehicle 161
ā¢ Traj. duration : [30, 2000] steps , endpts dist. > 50 pixels
āŖ TrajNet dataset:
ā¢ 58 scenes from UCY, ETH and SSD dataset
ā¢ 11,448 pedestrian traj.
ā¢ 20 steps each traj., world coordinates in meter.
32
![Page 33: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/33.jpg)
Implementation Details
āŖ Running on one RTX 2080 Ti GPU with Nvidia docker image
āŖ Optimization tricks:
ā¢ gradient clipping to 50% gradient norm
ā¢ Adam optimizer, lr = 0.005, lr decay to 50%
āŖ Dynamic length batches
āŖ Pre-computed features for accelerating training speed.
āŖ Training time:
ā¢ Voyager: 1 hr for 1000 epochs with 3 MDN output heads
ā¢ Trajnet: ~30 mins for 1700 epochs with 12 MDN output heads
33
![Page 34: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/34.jpg)
Experimental Results ā Voyager dataset
34
Group ID Method RMSE@10 RMSE@20 RMSE@40
Baselines
1 Linear Reg (š = 1) 62.47 68.59 82.51
2 VAR (š = 5) 46.85 90.27 163.02
3 MLP + Reg 14.17 27.08 50.16
4 LSTM+Reg 8.67 14.65 27.39
Ours
5 LSTM+MDN 7.42 13.26 25.256 LSTM+MDN (single output) 7.51 (0.22)* 13.30 (0.34) 25.20 (0.45)
7 LSTM+MDN+OccuMap 7.24 (0.02) 12.70 (0.008) 24.30 (0.01)
8 LSTM+MDN+Attribute 7.22 (0.0003) 12.95 (0.01) 24.74 (0.02)
9 LSTM+Traj. Feature 7.39 (0.03) 12.89 (0.05) 24.45 (0.03)
10 LSTM+MDN+OccuMap
+Attribute7.30 (0.09) 12.71 (0.005) 24.22 (0.004)
11 LSTM+MDN+OccuMap
+Attribute + Traj. Feature7.36 (0.04) 13.06 (0.03) 24.54 (0.008)
* p-values against method 5 (LSTM+MDN), p < 0.05 means two results are different with statistical significance
Experiment results and ablation study (error in pixels)
![Page 35: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/35.jpg)
Experimental Results ā TrajNet dataset
35
Group ID Method Average error Final error Mean error
Social
LSTM*
9 Occupancy LSTM 2.1105 3.12 1.101
10 Social LSTM 1.3865 2.098 0.675
Ours**
4 LSTM+Reg 1.039 1.382 0.696
5 LSTM+MDN 1.036 1.377 0.694
7 LSTM+MDN+OccuMap 1.028 1.370 0.686
*Unofficial Implementation from https://github.com/quancore/social-lstm
**cross validation result on train set because evaluation server not available
Tentative comparison between Social LSTM and Ours (error in meters)
![Page 36: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/36.jpg)
Qualitative Results ā Easy Example
36
Loc. at t t+10 t+20 t+40
Forecasted
Actual
Whole Traj.
x
y
![Page 37: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/37.jpg)
Qualitative Results ā Easy Example
37
Loc. at t t+10 t+20 t+40
Forecasted
Actual
Whole Traj.
x
y
![Page 38: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/38.jpg)
Qualitative Results - Intermediate Difficulty
38
Loc. at t t+10 t+20 t+40
Forecasted
Actual
Whole Traj.
x
y
![Page 39: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/39.jpg)
Qualitative Results - Intermediate Difficulty
39
Loc. at t t+10 t+20 t+40
Forecasted
Actual
Whole Traj.
x
y
![Page 40: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/40.jpg)
Qualitative Results ā Hard Example
40
Loc. at t t+10 t+20 t+40
Forecasted
Actual
Whole Traj.
x
y
![Page 41: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/41.jpg)
Object Detection
+Object Tracking
Input video
Task: Forecast Entering Excavation Zone Events
1. Using trajectory forecasting model to predict person/vehicleās future locations in 0.5/1.0/2.0 seconds
2. Matching predictions to human-defined excavation zones.
A Safety Application Prototype
41
Object detection + tracking:ā¢ Mask RCNN (Resnet-101 backbone, Caffe2 Model zoo) for Person & Vehicle ā¢ SORT for tracking Person & Vehicle objects
![Page 42: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/42.jpg)
Admin panel to modify regions of interest
A Safety Application Prototype
42
![Page 43: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/43.jpg)
Viewer panel
A Safety Application Prototype
43
![Page 44: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/44.jpg)
Demo Video
A Safety Application Prototype
44
![Page 45: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/45.jpg)
Summary
āŖ Improving construction safety requires more frequent, accurate and proactive inspections.
āŖ We show detection, tracking, and trajectory forecasting models are promising ways to improve predictive construction safety management.
45
![Page 46: Video-Based Activity Forecasting for Construction Safety ......Video-Based Activity Forecasting for Construction Safety Monitoring Use Cases Speaker: Shuai Tang University of Illinois](https://reader033.fdocuments.net/reader033/viewer/2022060913/60a7b56836f1e11037235d5c/html5/thumbnails/46.jpg)
Video-Based Activity Forecasting
for Construction Safety Monitoring Use Cases
Shuai [email protected]
GTC, Santa Jose, 2019