CNN AV1 Intra Encoder

• Motivation
• Evolution of Video Codec
• Encoder and Decoder
  • Encoder
  • Decoder
• AOM AV1
  • Partition Modes
  • Intra Prediction Tools
  • AV1 Performance Review
• Video Codec Acceleration
  • Parallelism
  • Statistic Based Approaches
  • Machine Learning Based Approaches
• Convolutional Neural Network
  • General Structure of CNN
  • Training CNN Model
• CNN AV1 Intra Encoder
  • Partition Decision for Intra Frame in AV1
  • Current Fast Partition Mode Selection Strategies in AV1
  • Partition Mode Prediction with CNN for Intra Frame
  • CNN Model in This Research
  • Encoder Modification
• Training CNN Model
  • Labels Extraction from the Encoder
  • Training and Evaluation Setting
• Encoding Efficiency Metrics
  • PSNR
  • BD-PSNR
  • BD-BR
• Performance Evaluation
  • Experimental Configuration
  • Performance of Reference AV1 Encoder Compared with HEVC
  • Analysis of Dataset
  • Evaluation of CNN models with different dataset
  • Performance of CNN Intra Encoder
• Conclusion and Future Perspective
  • Conclusion
  • Future Perspective
• Appendix
  • Video frames for model training
  • Video frames for test
  • Partitions Comparison Between Two Encoders
• Source Code
  • model 1
  • model 2
  • Encoder Modification
• Tools
• TEMP
  • Training results of None and Split partition modes only
  • Training results of Horz and Vert partition modes only
  • Training results of None and Horz partition modes only
  • Training results of Horz and Split partition modes only
  • Training results of None, Horz and Split partition modes only
  • Training results of Horz4 and Vert4 partition modes only
  • Training results of Split and Horz4 partition modes only
  • Training results of Split and Vert4 partition modes only
  • Training results of Split, Horz4 and Vert4 partition modes only
  • Training results of HorzA and HorzB partition modes only
  • Training results of None, Split and the rest
  • Comparison between seperate qp and mixed qps

Indices and tables

• Index
• Module Index
• Search Page