ELTrack: Correlating Events and Language for Visual Tracking
- This repository contains the official implementation of ELTrack: Correlating Events and Language for Visual Tracking.
Samples of VisEvent dataset where the first row represent the RGB-image, the second row represent the raw events-image representation adopted in VisEvent, and the third row is our proposed Pseudo-Frames events-image representation. (a) A vehicle in a tunnel scene; (b) A drone in front of a tall building; (c) A cat lying on a surface; (d) An indoor scene with a partially visible person; (e) A logo of letters next to a computer monitor with visible wiring or cables; (f) A scene with flowers; and (g) An outdoor tennis court or sports area.
-
Language description of each column in the figure above is given as: -(a) A view through a vehicle's windshield, showing a road ahead, with a focus on the driver's perspective and the car's dashboard.
- (b) A drone is captured in flight set against the contrasting backdrop of high-rise buildings bathed in bright light.
- (c) Domestic cat in seated position. A black cat sits with poise amidst a domestic setting. The scene is softly illuminated, revealing a juxtaposition of household appliances and a textured floor, hinting at the quietude of a home space.
- (d) A person in a dimly lit room is rotating a ball attached to a string, with a laptop on a desk and a door in the background.
- (e) Digits forming the letters 'AI' moving across the screen, changing position and orientation in each frame.
- (f) Flower in varying stages of bloom, captured in natural light.
- (g) Tennis player moving laterally on the court.