MLLM can recognize and track anything in videos now!

Multi-modal Large Language Models (MLLMs) have demonstrated their ability to perceive objects in still images, but their application in video-related tasks, such as object tracking, remains understudied. This lack of exploration is primarily due to two key challenges. Firstly, extensive pretraining on large-scale video datasets is required to equip MLLMs with the capability to perceive objects across multiple frames and understand inter-frame relationships. Secondly, processing a large number of frames within the context window of Large Language Models (LLMs) can impose a significant computational burden. To address the first challenge, we introduce ElysiumTrack-1M, a large-scale video dataset supported for three tasks: Single Object Tracking (SOT), Referring Single Object Tracking (RSOT), and Video Referring Expression Generation (Video-REG). ElysiumTrack-1M contains 1.27 million annotated video frames with corresponding object boxes and descriptions. Leveraging this dataset, we conduct training of MLLMs and propose a token-compression model T-Selector to tackle the second challenge. Our proposed approach, Elysium: Exploring Object-level Perception in Videos via MLLM, is an end-to-end trainable MLLM that attempts to conduct object-level tasks in videos without requiring any additional plug-in or expert models.

Referring Single Object Tracking (RSOT)

We use prompt "Please find {expression} in the initial frame and provide the detailed coordinates in each frame." for each video.

a running dog played in the snow field	the cap on a dog's head	the snow field	shoes	the person in red

boy back to camera	a dancing kangaroo	dog

Single Object Tracking (SOT)

We use prompt "This is a video showing an object with coordinates {coordinates} in Frame 1. Provide the detailed coordinates of the object in each frame." for each video.

[34,40,51,67]	[35,48,60,55]