VTTI / Event-Correlation

Inferencing for external driving events such as a lane change based on gaze estimation and object detection input using a dash cam and a face camera.

Event-Correlation

Inferencing for external driving events such as a lane change based on gaze estimation and object detection input using a dash cam and a face camera.

Required Files: Event Correlation

1.) All Epochs File

Format: CSV with the given columns

trialID	epochID	epochType	timeOfDay	baseballCap	glasses	epochBeginTimestamp	epochEndTimestamp	epochBeginFrame	epochEndFrame

2.) All processed CSV Files Dir

Real Values CSV:

e.g., dgf_CMVP_0000_0000_10_130204_1926_00031_Face_real_bbox.csv

Frame#		X-value for Gaze	Y-value for Gaze	Object Name

Adjusted to Windshield Camera CSV:

e.g., dgf_l2cs_CMVP_0000_0000_10_130218_1429_00079_Face_adjusted_bbox.csv

Frame#	X-value for Gaze	Y-value for Gaze	Object Name

3.) Mapping CSV file for Epochs to processed

Format: dataset.csv

trialID		CSV File Name (Detections and Gaze)

4.) Output directory for histogram images

For a demo, use the pre-existing folder on VTTI server: vtti/projects03/451600/Working/FromSandesh/Event Correlation Folders/

For a smaller sample set use the one provided with the repository.

5.) How to run

Installation:

pip3 install -r requirements.txt

For custom folders use the arguments

python3 event_correlation_hist.py --out_dir='' --epochs_path='' --mapped_path='' --detections_dir=''