Logical Scenario Matching from Video [Draft]:

Give a short introduction about what a logical scenario is.

• YOLO can detect car/truck/...
• Lane lines can be detected with OpenCV (maybe YOLO also? other CNN?)
• Can we assign absolute lane_id for objects? for ego vehicle? -- If so, we can label atomic states for moving objects and do the logical scenario matching
• Even this is not possible, still some info can be extracted from video. --- Lane changes? --- How many objects? --- Object type validation

Optional: - Run YOLO (OpenCV or TF, or whatever) with Spark on cluster and process videos

Actions: 1. Install OpenCV on Linux vm 2. Install/test darknet-yolo python bindings for single frames 3. Process a session video with darknet-yolo 4. Detect lane lines 5. Combine objects and lane lines

• Run OpenCV with Scala