論文まとめ

• githubのmarkdownプレビューだと数式が崩壊してしまうため、pdf参照

論文

• vision-basic
  • u-net : U-Net(Convolutional Networks for Biomedical Image Segmentation)
  • vision-transformer : VisionTransformer(AN IMAGE IS WORTH 16X16 WORDS: TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE)
  • swin-transformer : SwinTransformer(Swin Transformer: Hierarchical Vision Transformer using Shifted Windows)
  • distillation : Distilling the Knowledge in a Neural Network
  • maxvit : MaxViT: Multi-Axis Vision Transformer
  • mae : Masked Autoencoders Are Scalable Vision Learners
  • simmim : SimMIM: a Simple Framework for Masked Image Modeling
  • revnet : The Reversible Residual Network: Backpropagation Without Storing Activations
  • rev-vit : Reversible Vision Transformers
  • dino: Emerging Properties in Self-Supervised Vision Transformers
  • dinov2: DINOv2: Learning Robust Visual Features without Supervision
  • dinov3: DINOv3
• generation (diffusion, flow)
  • ddpm : Denoising Diffusion Probabilistc Models
  • palette : Palette: Image-to-Image Diffusion Models
  • ddim : Denoising Diffusion Implicit Models
  • improved-ddpm : Improved Denoising Diffusion Probabilistic Models
  • adm : Diffusion Models Beat GANs on Image Synthesis
  • glide : Guided Language to Image Diffusion for Generation and Editing
  • ldm : Latent Diffusion Model(Stable diffusion)
  • cdm : Cascaded Diffusion Model
  • inpaint-survey : Deep Learning-based Image and Video Inpainting: A Survey
  • dit: Scalable Diffusion Models with Transformers
  • qwen-image: Qwen-Image Technical Report
  • flow-matching: FLOW MATCHING FOR GENERATIVE MODELING
• super-resolution
  • srcnn-vdsr-fsrcnn-fspcn : 超解像の歴史(CNNあたりからGAN登場まで)
  • swinir : SwinIR(SwinIR: Image Restoration Using Swin Transformer)
  • hat : HAT-L(Hybrid Attention Transformer)
  • drct : DRCT(Dense Residual Connected Transformer)
  • sr3 : Image Super-Resolution via Iterative Refinement
  • ipg : Image Processing GNN: Breaking Rigidity in Super-Resolution
  • yonos-sr : You Only Need One Step: Fast Super-Resolution with Stable Diffusion via Scale Distillation
  • hmanet : HMANet: Hybrid Multi-Axis Aggregation Network for Image Super-Resolution
  • diffusion-sr-survey : Diffusion Models, Image Super-Resolution
  • blip-diffusion: BLIP-Diffusion: Pre-trained Subject Representation for Controllable Text-to-Image Generation and Editing
  • controlnet: Adding Conditional Control to Text-to-Image Diffusion Models
  • prompt-to-prompt: Prompt-to-Prompt Image Editing with Cross Attention Control And Everything: A Survey
  • tr-misr : TR-MISR: Multiimage Super-Resolution Based on Feature Fusion With Transformers
  • div2k : NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study
  • lsdir : LSDIR: A Large Scale Dataset for Image Restoration
  • df2k : DF2K
  • ntire-challenge-on-lfsr : NTIRE 2024 Challenge on Light Field Image Super-Resolution: Methods and Results
  • epit : (EPIT)Learning Non-Local Spatial-Angular Correlation for Light Field Image Super-Resolution
  • pixel-shuffle : Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network
  • datsr : Reference-based Image Super-Resolution with Deformable Attention Transformer
  • ais2024challenge-survey : Real-Time 4K Super-Resolution of Compressed AVIF Images. AIS 2024 Challenge Survey
• image-restoration
  • aioir-survey : A Survey on All-in-One Image Restoration: Taxonomy, Evaluation and Future Trends
  • ram : Restore Anything with Masks: Leveraging Mask Image Modeling for Blind All-in-One Image Restoration
• deblurring
  • image-deblurring-survey : Deep Image Deblurring: A Survey
  • adarevd : AdaRevD: Adaptive Patch Exiting Reversible Decoder Pushes the Limit of Image Deblurring
• object-detecion
  • faster-r-cnn: Faster R-CNN: Towards Real-Time Object Detection with Reg
  • yolo: You Only Look Once: Unified, Real-Time Object Detection
  • yolov2: YOLO9000: Better, Faster, Stronger
  • yolov3: YOLOv3: An Incremental Improvement
  • yolov4: YOLOv4: Optimal Speed and Accuracy of Object Detection
  • yolox: YOLOX: Exceeding YOLO Series in 2021
  • detr: End-to-End Object Detection with Transformers
  • ovd-survey: A Survey on Open-Vocabulary Detection and Segmentation: Past, Present, and Future
  • ovr-cnn: Open-Vocabulary Object Detection Using Captions
  • vild: OPEN-VOCABULARY OBJECT DETECTION VIA VISION AND LANGUAGE KNOWLEDGE DISTILLATION
  • mrvg: Multimodal Reference Visual Grounding
  • nids: Adapting Pre-Trained Vision Models for Novel Instance Detection and Segmentation
• segmentation
  • sam: Segment Anything
  • sam2: SAM 2: Segment Anything in Images and Videos
  • sam3: SEGMENT ANYTHING WITH CONCEPTS
• 3dgs
  • 3dgs : 3D Gaussian Splatting for Real-Time Radiance Field Rendering
  • srgs: SRGS: Super-Resolution 3D Gaussian Splatting
  • gaussiansr : GaussianSR: 3D Gaussian Super-Resolution with 2D
  • supergaussian : SuperGaussian: Repurposing Video Models for 3D Super Resolution Diffusion Priors
  • supergs : SuperGS: Super-Resolution 3D Gaussian Splatting via Latent Feature Field and Gradient-guided Splitting
  • e-3dgs : Per-Gaussian Embedding-Based Deformation for Deformable 3D Gaussian Splatting
  • deblurring-3dgs : Deblurring 3D Gaussian Splatting
• nerf
  • nerf : NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
  • nerf-sr : NeRF-SR: High Quality Neural Radiance Fields using Supersampling
  • mip-nerf : Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields
  • crop : Cross-Guided Optimization of Radiance Fields with Multi-View Image Super-Resolution for High-Resolution Novel View Synthesis
• video
  • adatad : End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames
  • iaw : Aligning Step-by-Step Instructional Diagrams to Video Demonstrations
  • vtune: On the Consistency of Video Large Language Models in Temporal Comprehension
  • longvale: LongVALE: Vision-Audio-Language-Event Benchmark TowardsTime-Aware Omni-Modal Perception of Long Videos
  • video-3d-llm: Video-3D LLM: Learning Position-Aware Video Representation for 3D Scene Understanding
  • m-llm-based-video-frame-selection: M-LLM Based Video Frame Selection for Efficient Video Understanding
  • video-comp: VideoComp: Advancing Fine-Grained Compositional and Temporal Alignment in Video-Text Models
  • seq2time: Seq2Time: Sequential Knowledge Transfer for Video LLM Temporal Grounding
  • video-models-are-zero-shot-learners-and-reasoners: Video models are zero-shot learners and reasoners
• vision-and-language
  • clip : CLIP(Learning Transferable Visual Models From Natural Language Supervision)
  • lit : LiT : Zero-Shot Transfer with Locked-image text Tuning
  • blip : BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
  • blip2 : BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models
  • siglip : Sigmoid Loss for Language Image Pre-Training
  • Flamingo: a Visual Language Model for Few-Shot Learning
  • video-llm-survey : Video Understanding with Large Language Models: A Survey(途中)
  • llava : Visual Instruction Tuning
  • llava-next-video : blog
  • llava-next-stronger : blog
  • llava-video : VIDEO INSTRUCTION TUNING WITH SYNTHETIC DATA
  • llava-next-interleave : LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models
  • long-vlm : LongVLM: Efficient Long Video Understanding via Large Language Models(ECCV2024)
  • tcr : Text-Conditioned Resampler For Long Form Video Understanding(ECCV2024)
  • qwen-vl : Qwen2.5-VL Technical Report
  • vtg-llm : VTG-LLM: Integrating Timestamp Knowledge into Video LLMs for Enhanced Video Temporal Grounding
  • internvl : InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks
  • internvl-1_5 : How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites
  • internvl-3 : InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
  • video-xl : Video-XL: Extra-Long Vision Language Model for Hour-Scale Video Understanding
  • video-xl-pro : Reconstructive Token Compression for Extremely Long Video Understanding
  • phi3-tech-report : Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone
  • phi4-mini : Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs
  • deepseek-vl : DeepSeek-VL: Towards Real-World Vision-Language Understanding
  • deepseek-vl2 : DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding
  • vidi: Vidi: Large Multimodal Models for Video Understanding and Editing
  • ref-l4: Revisiting Referring Expression Comprehension Evaluation in the Era of Large Multimodal Models
  • molmo-and-pixmo: Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models
  • llava-st: LLaVA-ST: A Multimodal Large Language Model for Fine-Grained Spatial-Temporal Understanding
  • internvl-3_5: InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
  • vlm-vg: Learning Visual Grounding from Generative Vision and Language Model
  • streaming-vlm: STREAMINGVLM: REAL-TIME UNDERSTANDING
  • step3-vl: STEP3-VL-10B Technical Report FOR INFINITE VIDEO STREAMS
  • piza: Referring Expression Comprehension for Small Objects
  • paddle-ocr: PaddleOCR-VL: Boosting Multilingual Document Parsing via a 0.9B Ultra-Compact Vision-Language Model
  • deepseek-ocr: DeepSeek-OCR: Contexts Optical Compression
  • idefic: Building and better understanding vision-language models: insights and future directions
  • covt: Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens
  • mcot: Multimodal Chain-of-Thought Reasoning in Language Models
  • viscot: visual CoT: Advancing Multi-Modal Language Models with a Comprehensive Dataset and Benchmark for Chain-of-Thought Reasoning
  • qwen3-vl-embedding-reranker: Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking
  • migician: Migician: Revealing the Magic of Free-Form Multi-Image Grounding in Multimodal Large Language Models
• mi (Mechanistic Interpretability)
  • survey-multimodal-mi: A Survey on Mechanistic Interpretability for Multi-Modal Foundation Models
  • latentlens: LATENTLENS: Revealing Highly Interpretable Visual Tokens in LLMs
  • attention-lens-tool: Attention Lens: A Tool for Mechanistically Interpreting the Attention Head Information Retrieval Mechanism
  • vlm-attention-lens: Devils in Middle Layers of Large Vision-Language Models: Interpreting, Detecting and Mitigating Object Hallucinations via Attention Lens
• nlp
  • transformer : Transformer(Attention is all you need)
  • perceiver: Perceiver: General Perception with Iterative Attention
  • perceiver: PERCEIVER IO: A GENERAL ARCHITECTURE FOR STRUCTURED INPUTS & OUTPUTS
  • lora : LORA: LOW-RANK ADAPTATION OF LARGE LANGUAGE MODELS
  • auxiliary-loss-free : auxiliary-loss-free load balancing strategy for mixture-of-experts
  • deepseek-v3 : DeepSeek-V3 Technical Report
  • rag: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
  • prompt-tuning: The Power of Scale for Parameter-Efficient Prompt Tuning
  • norm-based-analysis: Attention is Not Only a Weight: Analyzing Transformers with Vector Norms