Download two RGB features extracted from Kinetics-I3D-PyTorch.
The annotation file in the repository (dataset/annotation-Mar9th-25fps.pkl) contains the list for causality annotation each video and its meta information.
- Each element in the list has video meta information and cause and effect event labels.
- traffic accident video information
- (v_Youtube clip ID, start time in Youtube clip, end time in Youtube clip)
- cause annotation
- (cause semantic label, cause start time, cause end time, cause semantic label index)
- effect annotation
- (effect semantic label, effect start time, effect end time, effect semantic label index)
- traffic accident video information
Note that removing the prefix v_ to search a video on youtube and all time stamps are written in seconds.
We have 17 and 7 semantic labels for cause and effect event correspondingly.
- For cause labels, we adopt semantic taxonomy introduced in the crash avoidance research. The research introduced a new typology of pre-crash scenario of traffic accident. The typology of pre-crash serves as a semantic taxonomy of cause events in traffic accident. We merge labels With Prior Vehicle Action and Without Prior Vehicle Action into the same labels because it is hard to be dicriminated by only watching video in many traffic accidents.
- For effect event, we use 7 semantic labels which frequently appeared in collected videos with traffic accident.
- The prior distributions of both cause and effect event can be calculated by aggregating ocurrences of individual cause and effect events in the research, which is shown in figure 4 of the paper.
- We modify BeaverDam to support both temporal regions and spatio-temporal regions of cause and effect event.
- But, we annotate videos with temporal localization due to an expensive annotation cost and the ambiguity of cause event of accident in spatio-temporal regions.