mlsanityData to 7 Oct 2026

BlockCopy: High-Resolution Video Processing with Block-Sparse Feature Propagation and Online Policies

arXiv 2108.09376 ↗First posted 20 Aug 2021cs.CVcited by 9

Not on the map. Why?

Abstract (arXiv)

In this paper we propose BlockCopy, a scheme that accelerates pretrained frame-based CNNs to process video more efficiently, compared to standard frame-by-frame processing. To this end, a lightweight policy network determines important regions in an image, and operations are applied on selected regions only, using custom block-sparse convolutions. Features of non-selected regions are simply copied from the preceding frame, reducing the number of computations and latency. The execution policy is trained using reinforcement learning in an online fashion without requiring ground truth annotations. Our universal framework is demonstrated on dense prediction tasks such as pedestrian detection, instance segmentation and semantic segmentation, using both state of the art (Center and Scale Predictor, MGAN, SwiftNet) and standard baseline networks (Mask-RCNN, DeepLabV3+). BlockCopy achieves significant FLOPS savings and inference speedup with minimal impact on accuracy.

Key facts. Cited by 9 papers in the map (data through 7 Oct 2026). Builds on 41 of its 50 references that are in the map. First posted on arXiv 20 Aug 2021.

Builds on

41 of 50 references.

References · 41

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  4. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  5. 2015Distilling the Knowledge in a Neural Networkcited by 10,076
  6. 2017Mask R-CNNcited by 7,259
  7. 2016The Cityscapes Dataset for Semantic Urban Scene Understandingcited by 4,427
  8. 2018Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentationcited by 3,210
  9. 2017Non-local Neural Networkscited by 2,290
  10. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
  11. 2017A Closer Look at Spatiotemporal Convolutions for Action Recognitioncited by 1,139
  12. 2016Pruning Filters for Efficient ConvNetscited by 1,129
29 more
  1. 2016Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activationscited by 413
  2. 2017SkipNet: Learning Dynamic Routing in Convolutional Networkscited by 328
  3. 2016Deep Feature Flow for Video Recognitioncited by 186
  4. 2017BlockDrop: Dynamic Inference Paths in Residual Networkscited by 185
  5. 2017CityPersons: A Diverse Dataset for Pedestrian Detectioncited by 171
  6. 2017Convolutional Networks with Adaptive Inference Graphscited by 147
  7. 2020Sparse GPU Kernels for Deep Learningcited by 138
  8. 2016Spatially Adaptive Computation Time for Residual Networkscited by 118
  9. 2019In Defense of Pre-trained ImageNet Architectures for Real-time Semantic Segmentation of Road-driving Imagescited by 99
  10. 2019Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inferencecited by 91
  11. 2018Accel: A Corrective Fusion Network for Efficient Semantic Segmentation on Videocited by 78
  12. 2016Semantic Video Segmentation by Gated Recurrent Flow Propagationcited by 76
  13. 2016Seq-NMS for Video Object Detectioncited by 68
  14. 2019Adaptive NMS: Refining Pedestrian Detection in a Crowdcited by 66
  15. 2017Semantic Video CNNs through Representation Warpingcited by 65
  16. 2020Efficient Semantic Video Segmentation with Per-frame Inferencecited by 63
  17. 2018Low-Latency Video Semantic Segmentationcited by 61
  18. 2018Online Model Distillation for Efficient Video Inferencecited by 60
  19. 2018Occlusion-aware R-CNN: Detecting Pedestrians in a Crowdcited by 58
  20. 2016Clockwork Convnets for Video Semantic Segmentationcited by 55
  21. 2018SBNet: Sparse Blocks Network for Fast Inferencecited by 53
  22. 2018Dynamic Video Segmentation Networkcited by 51
  23. 2020Spatially Adaptive Inference with Stochastic Feature Sampling and Interpolationcited by 51
  24. 2019Mask-Guided Attention Network for Occluded Pedestrian Detectioncited by 36
  25. 2017Temporal 3D ConvNets: New Architecture and Transfer Learning for Video Classificationcited by 27
  26. 2019Looking Fast and Slow: Memory-Guided Mobile Video Object Detectioncited by 22
  27. 2019Efficient Video Semantic Segmentation with Labels Propagation and Refinementcited by 18
  28. 2019Semantic Head Enhanced Pedestrian Detection in a Crowdcited by 8
  29. 2018Fast Semantic Segmentation on Video Using Block Motion-Based Feature Interpolationcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 2; 2023: 2; 2024: 3; 2025: 1; 2026: 1.

9 papers.

  1. 2022Single UHD Image Dehazing via Interpretable Pyramid Networkcited by 13
  2. 2024DSRC: Learning Density-insensitive and Semantic-aware Collaborative Representation against Corruptionscited by 13
  3. 2023Efficient Semantic Segmentation by Altering Resolutions for Compressed Videoscited by 8
  4. 2022Edge Video Analytics: A Survey on Applications, Systems and Enabling Techniquescited by 7
  5. 2024Multi-resolution Rescored ByteTrack for Video Object Detection on Ultra-low-power Embedded Systemscited by 4
  6. 2023Learning Online Policies for Person Tracking in Multi-View Environmentscited by 1
  7. 2024Multi-Exposure Image Fusion via Distilled 3D LUT Grid with Editable Modecited by 0
  8. 2025Leveraging Motion Estimation for Efficient Bayer-Domain Computer Visioncited by 0
  9. 2026MR2-ByteTrack: CNN and Transformer-based Video Object Detection for AI-augmented Embedded Vision Sensor Nodescited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025SDAR-VL: Stable and Efficient Block-wise Diffusion for Vision-Language Understanding
  2. 2026DFSAttn: Dynamic Fine-grained Sparse Attention for Efficient Video Generation
  3. 2026MVP: Enhancing Video Large Language Models via Self-supervised Masked Video Prediction
  4. 2026LVSA: Training-Free Sparse Attention for Long Video Diffusion
  5. 2026Sparse Forcing: Native Trainable Sparse Attention for Real-time Autoregressive Diffusion Video Generation
  6. 2025VSA: Faster Video Diffusion with Trainable Sparse Attention
  7. 2024Block Selective Reprogramming for On-device Training of Vision Transformers
  8. 2022Diffusion Models for Video Prediction and Infilling