541 technical terms and definitions
video action, localization temporal grounding, nms passed, video action localization & temporal grounding, video action localization, action localization temporal, action localization
Video Action Localization & Temporal Grounding Introduction & Motivation Video Action Localization: localize when actions occur in videos. Temporal boundaries, action recognition.…
Video Action Recognition with Spatiotemporal Transformers Introduction & Motivation Video Action Recognition with Spatiotemporal Transformers addresses a central problem in vision…
computer vision
Video Captioning is the task of automatically generating a natural language summary of a video clip — requiring the model to process spatiotemporal information (motion, audio…
video captioning description, video captioning, captioning description
Video Captioning Description Introduction & Motivation Video Captioning Description addresses a central problem in vision process control and video-description-caption-generation…
Video Captioning Generation Introduction & Motivation Video Captioning Generation addresses a central problem in multimodal learning and caption-accuracy-relevance alignment: how…
video generation
Video captioning models are the multimodal systems that convert temporal visual content into coherent natural language descriptions - they must summarize objects, actions…
Video Clip Retrieval from Text Introduction & Motivation Video Clip Retrieval from Text addresses a central problem in multimodal learning and temporal-relevance-accuracy…
hevc avs3 av1 hardware encoder, video encode decode asic, codec pipeline architecture, cabac entropy coding chip
Video Codec Chip Design: H.265/H.266 Hardware Encoder/Decoder — specialized ASIC for efficient video compression supporting 8K HDR streaming with <10 pJ/bit power efficiency Video…
video generation
Video completion is the broader temporal generation task of reconstructing large missing spatiotemporal regions, not just small holes in single frames - it often requires…
video generation
Video deblurring is the restoration task that removes motion blur by combining temporal evidence from adjacent frames and recovering high-frequency detail - it addresses blur…
Video Deblurring and Motion Compensation Introduction & Motivation Video Deblurring and Motion Compensation addresses a central problem in vision processing…
video generation
Video denoising is the process of removing random noise from frame sequences by combining spatial priors with temporally aligned multi-frame evidence - it improves signal-to-noise…
multimodal ai
Video Diffusion is a diffusion-based approach that generates or edits videos through iterative denoising over spatiotemporal representations - It offers high-quality motion…
video generation
Video diffusion models is the generative models that extend diffusion processes to produce coherent sequences of frames over time - they model both visual quality per frame…
video generation
Video editing with diffusion is the video transformation approach that applies diffusion-based generation to modify style, objects, or attributes across frames - it brings…
Video Frame Interpolation Introduction & Motivation Video Frame Interpolation addresses a central problem in vision processing and interpolation-smoothness prediction: how…
ai video generation, text to video, image to video, video diffusion, sora, runway, pika, stable video diffusion
Video generation synthesizes temporal visual sequences from text, images, keyframes, masks, motion controls, audio, or source video. It extends generative media into space-time…
generative models
Video generation creates video sequences from various input modalities — text descriptions, single images, sketches, or other videos — representing one of the most challenging…
multimodal ai
Video Generation is synthesizing coherent video sequences from learned generative models conditioned on prompts or context - It extends image generation to temporal content…
video understanding
Video grounding is the problem of locating the exact temporal region in a video that corresponds to a text query - given natural language such as an action phrase, the model…
video highlight detection, video highlight, highlight detection
Video Highlight Detection Introduction & Motivation Video Highlight Detection addresses a central problem in vision process control and highlight-detection-interesting-moment…
video generation
Video inpainting is the task of filling missing or masked regions in video frames while preserving spatial realism and temporal continuity - it combines reconstruction, motion…
multimodal ai
Video Inpainting is filling missing or corrupted regions in videos while preserving temporal and semantic consistency - It restores damaged footage and enables object removal…
Video Inpainting and Completion Introduction & Motivation Video Inpainting and Completion addresses a central problem in vision processing and inpainting-temporal-consistency…
video understanding
Video instance segmentation (VIS) is the task of segmenting each object instance in every frame while maintaining consistent identity across time - it unifies detection…
multimodal ai
Video-language pre-training is the multimodal learning paradigm that aligns video representations with textual descriptions such as narration, captions, or transcripts - it…
Video-Language Understanding Introduction & Motivation Video-Language Understanding addresses a central problem in multimodal learning and video-text-alignment-quality alignment…
video generation
Video Matting is a highly advanced, pixel-precise computer vision task that extracts a continuous alpha matte (transparency map with values ranging from to ) for foreground…
video generation
Video object removal is the editing task that removes selected objects from a video and reconstructs believable background content across all affected frames - it combines…
computer vision
Video Object Segmentation (VOS) is a dense prediction task that assigns a pixel-level mask to specific objects in every frame of a video — requiring the model to propagate…
video object segmentation vos, video object segmentation, object segmentation vos, video object, object segmentation, segmentation vos
Video Object Segmentation VOS Introduction & Motivation Video Object Segmentation VOS addresses a central problem in vision process control and object-segmentation-video-temporal…
video object tracking detection, video object tracking, object tracking detection, video object, object tracking, tracking detection
Video Object Tracking Detection Introduction & Motivation Video Object Tracking Detection addresses a central problem in vision process control and tracking-temporal-consistency…
Video Object Tracking Methods Introduction & Motivation Video Object Tracking Methods addresses a central problem in vision processing and tracking-continuity prediction: how…
video understanding
Video panoptic segmentation is the unified dense prediction task that assigns every pixel either a trackable thing instance or a semantic stuff class across time - it extends…
video generation
Video prediction is the sequence modeling task that forecasts future frames from past frames to learn temporal dynamics and scene evolution - this objective can teach motion…
multimodal ai
Video Prediction is forecasting future frames from observed video context using learned dynamics models - It supports planning, simulation, and anticipatory generation tasks. What…
Video Frame Prediction Introduction & Motivation Video Frame Prediction addresses a central problem in generative generation and future-frame-accuracy synthesis: how to Design…
video quality assessment vqa, video quality assessment, quality assessment vqa, video quality, quality assessment, assessment vqa
Video Quality Assessment VQA Introduction & Motivation Video Quality Assessment VQA addresses a central problem in vision process control and video-quality-metric-assessment…
computer vision
Video Question Answering (VideoQA) is a multimodal task where a model answers natural language questions about a video clip — requiring the integration of visual features…
video understanding
Advanced video question answering is the task of answering natural language questions about events, objects, and causal relationships in a video stream - unlike simple…
Video Question Answering Introduction & Motivation Video Question Answering addresses a central problem in multimodal learning and qa-accuracy-comprehension alignment: how…
rag
Video retrieval is the retrieval method that locates relevant videos or time segments using transcript, frame, and metadata signals - segment-level retrieval is key for long…
Video Retrieval by Text Description Introduction & Motivation Video Retrieval by Text Description addresses a central problem in multimodal learning and retrieval-ranking-accuracy…
video generation
Video stabilization is the process of removing unwanted camera shake by estimating motion trajectory and warping frames to a smoothed path - it improves visual comfort…
computer vision
Video style transfer is the technique of applying artistic or photographic styles consistently across video frames — extending image style transfer to temporal sequences while…
video generation
Video stylization is the process of transforming video appearance to a target artistic or visual style while maintaining temporal coherence - it extends image style transfer…
video summarization key frame, video summarization key, summarization key frame, video summarization, summarization key, key frame
Video Summarization Key Frame Introduction & Motivation Video Summarization Key Frame addresses a central problem in vision process control and video-summarization-key-selection…
Text-Guided Video Summarization Introduction & Motivation Text-Guided Video Summarization addresses a central problem in multimodal learning and summary-relevance-quality…
video generation
Video super-resolution (VSR) is the process of reconstructing high-resolution frames from low-resolution video by exploiting temporal redundancy across neighboring frames - unlike…
multimodal ai
Video Super-Resolution is increasing video resolution while preserving temporal coherence across frames - It enhances detail without introducing frame-to-frame instability. What…