TSS 100

The Traffic Situation Similarity Dataset contains almost 7000 video clips sampled from the Berkeley DeepDrive Dataset. Each sample consists of a base video clip and 6 candidate clips. The candidate clips have been ranked by human evaluators from most similar to the base clip to least similar. The clips differ in length, with the shortest being 1 second and the longest being 40 seconds.

Data Format

The file TSS.csv records the samples. Each row is a single sample with a base clip and 6 comparison clips. Each cell contains a number that corresponds to the clip with that number as its file name: e.g., in sample 0 the base clip is the video 15096.mp4. The candidate clips have been labeled as to their order: "MostSimilar" is the clip most similar to the base clip, "2ndSimilar" is the second most similar, and so on. In sample 0 the most similar clip is 14293.mp4 and the second most similar clip is 23555.mp4.