Tiêu đề Project Foxglove Guidelines - Long Video Caption eval set creation .docx ✅

- In general, you can vary caption length based on video length as well as density of information in the video (~100-500 words/min), but should make sure to capture as much information as described below. Overview Summary: Caption should have an overall description describing key events, actions, subjects and objects in the video. Shouldn’t miss/mistake/hallucinate important information. A summary that provides a high-level description of the whole video, including background elements, main objects, secondary objects, camera motion, readable text, and activity category. When describing people, describe approximate, gender, age descriptions, clothing (including color), occupation/role (if applicable), pose, accessories. Avoids information about a person's racial or ethnic origin (including skin color, hair color, apparent nationality or citizenship), sexual orientation, political affiliation, health condition or disability, religion, membership in a trade union, facial features, or hair color. For videos with multiple objects or complex scenes, the caption must include an overview of the most prominent elements. It should also describe the main events, actions performed by the main characters, along with any peculiarities that make the video unique. Avoid using phrases like “The video has/shows ….”, directly describe what you are seeing. Two English bulldogs walking outdoors on a grass surface near a brick building. exploring the outdoors. Their features are distinct - one brown with white patches, one black with white patches. The black and white dog briefly pauses to sniff the base of the brick wall. The white and brown dog runs past the white and black dog and runs towards the camera licking it’s nose once. In the background, a field of green grass stretches out, with trees on the horizon. During the video, the camera pans to the right, briefly offering a glimpse of the leftmost side of a white window shutter on the brick wall.