Pith. sign in

REVIEW 3 major objections 5 minor 50 references

RL-RC-DoT: A Block-level RL agent for Task-Aware Video Compression

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A PPO-trained agent that adjusts per-block quantization in a standard H.264 encoder cuts the bit-rate needed for the same car-detection precision by about 25%.

desk verdict Genuinely new and practically useful RL-based QP control; the 25% BD-rate gains are real but conditional on a filtered dataset, so the paper needs a serious revision, not a desk reject. read the letter →

arxiv 2501.12216 v2 pith:H3IQQS4V submitted 2025-01-21 cs.LG cs.CVeess.IV

Uri Gadot , Assaf Shocher , Shie Mannor , Gal Chechik , Assaf Hallak This is my paper · ORCID
classification cs.LGcs.CVeess.IV
keywords videocompressiontask-awareencodingreinforcementlearningquantizationparametercontrolrate-distortionoptimizationobjectdetectionsaliency
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that a reinforcement-learning agent can steer a standard video encoder to favor whatever an AI downstream task needs, without retraining the encoder or running the task at encoding time. The agent, trained with PPO, picks a per-macroblock quantization-parameter delta for every frame of an H.264/x264 stream, using only statistics the encoder already computes. On a filtered subset of driving videos, it reports 24.7% lower bit-rate for the same car-detection precision than standard x264, at a 1.19% PSNR penalty, and similar gains for saliency-weighted quality. The policy transfers to other detectors and to segmentation without retraining. If correct, this points to a practical way to make existing codecs task-aware for edge and streaming use.

What carries the argument

The engine is a PPO agent (proximal policy optimization, a standard reinforcement-learning algorithm) that maps cheap per-frame encoder statistics—per-macroblock energy, intra and propagating costs, inverse quantization scale, plus global rate and frame-type features—to a low-resolution delta-QP map that is upsampled to all macroblocks of the frame. The reward is a weighted sum of a bit-rate term and a task score, where the task score compares the downstream model's output on the reconstructed frame against its output on the raw frame. A block-wise auxiliary loss predicts local reward components, giving the policy a denser learning signal. This machinery lets a standard x264 encoder be steered frame by frame, with the policy acting over the temporal sequence so it can weigh future bit-rate consequences.

What would settle it

Re-run the evaluation on the BDD100K streams that the paper filtered out because they showed zero car-detection precision at most target bit-rates; if RL-RC-DoT does not improve or worsens the detection BD-rate there, the central claim that per-block QP control helps task performance on arbitrary driving video fails.

Watch

Extended reading notes

Core claim

The central discovery is that per-block QP control by a policy trained to maximize a downstream task score plus a bit-rate penalty can reallocate bits toward task-relevant regions without needing task ground truth during inference. Training uses the raw frame as pseudo-ground truth: the task model's output on the uncompressed frame defines what the reconstruction should preserve. The same compressed stream stays watchable, because the PSNR cost is small. The paper demonstrates this for car detection and saliency-weighted PSNR, with BD-rate reductions of about 25% on each, and shows the learned policy improves performance of unseen detectors and a segmentation model.

Load-bearing premise

The evaluation only includes driving streams where the car detector already finds cars in the uncompressed video, so the reported savings may not carry over to streams where detection starts from zero.

Editorial extensions

If this is right

  • Encoding can be made task-aware on top of existing H.264 codecs, so deployment only adds a lightweight network that outputs QP deltas rather than a new codec stack.
  • At equal bit-rate, car-detection precision and recall improve; at equal precision, bit-rate drops by roughly 25%, with human-viewability nearly unchanged.
  • A policy trained for one detector improves a different detector and a segmentation model, so the compressed video remains reusable across models and tasks.
  • ROI and saliency encoding get similar bit-rate savings, suggesting the mechanism generalizes beyond detection.
  • The myopic ablation shows temporal lookahead matters; optimizing frame by frame loses most of the benefit, so the video-level formulation is essential.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The authors leave implicit that the dataset filter means the reported gains live in a regime where the detector already works; on streams with zero baseline precision the mechanism has no signal to reallocate toward.
  • A natural extension is to run the same reward scheme on HEVC or VVC by replacing per-macroblock statistics with coding-tree-unit features; the codec-agnostic formulation should transfer, though the action-space resolution may need retuning.
  • The block-wise reward-prediction head could double as a learned importance map, which might explain the 3.2% tracking BD-rate gain and could be used to seed ROI encoding without a saliency model.
  • Since the pseudo-ground-truth reward only needs a model output on raw and reconstructed frames, the method should also work with task-agnostic or learned perceptual metrics, reducing dependence on a specific detector during training.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes RL-RC-DoT, a PPO-based reinforcement learning agent that controls per-macroblock QP deltas in the x264 encoder to optimize a downstream video-analysis task rather than pure reconstruction quality. The MDP is defined at frame granularity: the state is a set of encoder statistics, the action is a low-resolution delta-QP map that is upsampled to the macroblock grid, and the reward combines a bit-rate fidelity term with a task term that compares the downstream model's output on the reconstructed frame against its output on the raw frame, used as pseudo-ground-truth. The method is evaluated on a filtered subset of BDD100K for car detection (YOLOv5-nano) and saliency-weighted PSNR (ROI encoding), with reported BD-rate reductions of about 24.7% and 25.64% relative to standard x264, small PSNR loss, positive transfer to SSD and segmentation, and ablations of the reward-prediction auxiliary loss and discount factor.

Significance. If the reported results are representative, RL-RC-DoT is a practically valuable contribution: it wraps a standard, widely deployed encoder with a real-time policy (claimed 250 FPS overhead), requires no downstream model or ground truth at inference time, and shows some robustness across tasks and detectors. The paper's strengths include the concrete system design, the use of a held-out test set, the named stream lists for reproducibility, the ablation studies, and the transfer experiments to SSD and segmentation. The main significance is limited by two correctness-risk concerns: the evaluation is restricted to a filtered set of streams with non-trivial baseline detection performance, and the primary car-detection metric is computed against a pseudo-ground-truth derived from the same detector used in the reward.

major comments (3)
  1. [Section 4.1] The dataset construction filters out streams that exhibit zero detection precision across most target bit-rates before the train/validation/test split, so all 100 test streams come from a pool where standard x264 encoding already yields non-trivial YOLOv5 precision. Consequently, the headline BD-rate numbers in Table 2 are conditional on this favorable subset, and the abstract/conclusion statements of general improvement over task-agnostic encoding are not supported for the unfiltered deployment distribution. Please report how many BDD100K streams were excluded, characterize the excluded streams, and provide an evaluation on the unfiltered set or at least on a held-out stratum with low baseline precision. Without such evidence the external validity of the central claim remains undetermined.
  2. [Sections 3.1 and 5.1] The car-detection metric is precision computed with YOLOv5-nano detections on the raw frames as pseudo-ground-truth, and the same pseudo-ground-truth is used in the training reward. This creates a genuine risk that the policy learns to make compressed-frame detections agree with raw-frame detections rather than to improve true detection quality against human annotations. Since BDD100K provides manual bounding-box annotations, please report AP/mAP and recall against those annotations as the primary task metric, and additionally show the correlation between the pseudo-GT-based and GT-based rankings. The SSD and segmentation transfer results partially mitigate this concern, but they do not by themselves validate the headline car-detection BD-rate numbers.
  3. [Sections 5.1 and 5.3] The paper reports the headline car-detection BD-rate with inconsistent uncertainty values: Section 5.1 states 24.7% (±1.38%), while Table 2 and Appendix Table 7 report −24.7 ± 1.57. Please reconcile these values and state explicitly the number of streams, rate points, and integration interval used for each BD-rate computation, since BD-rate is sensitive to these choices and to the filtering criterion.
minor comments (5)
  1. [Section 4.1] The paper does not state how many total BDD100K streams were considered or how many were removed by the filtering criterion; please add these counts so the reader can judge the selection effect.
  2. [Appendix 7.3.2, Table 9] The entry for the 'fast' preset reads '-22/8', which is almost certainly a typo for '-22.8'; please fix it.
  3. [Appendix 7.3.1] The additional downstream-task results (DAVIS segmentation and ByteTrack tracking) are reported without sufficient experimental detail; please specify which policy was used, what was matched at each bit-rate, and how the metrics were computed, or omit these results from the claims.
  4. [Appendix 7.7] The reproducibility section describes modified x264 code but provides no link to the code or model weights; releasing these would materially strengthen the reproducibility claim.
  5. [Section 6 and Abstract] The abstract and conclusion state improvements over task-agnostic encoding without noting that the evaluation was performed on a filtered subset of streams; please add the caveat or present the unfiltered results.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: policy training and held-out evaluation are separated, transfer results are external, and the dataset filter is a generality caveat rather than a circular step.

full rationale

The derivation chain is not circular. The RL policy is trained with a PPO objective whose reward combines bit-rate error with r_task = D[f_task(frameraw), f_task(framerecon)] (Section 3.1), where the raw-frame detector output serves as pseudo-ground-truth. The headline BD-rate results are computed on 100 held-out test streams (Section 4.1), with a separate validation set used only for hyperparameter selection; no test-stream rewards or test labels enter training. The reward is a consistency measure against raw-frame detections, while the reported precision/recall BD-rates are standard task metrics evaluated on the test set, so the evaluation is not the same quantity as the training reward. The most plausible near-circular concern—that optimizing a YOLOv5-nano-based reward guarantees YOLOv5-nano precision gains—is empirically countered by the transfer results in Table 2: SSD precision BD-rate of -26.2% and segmentation IoU BD-rate of -14.6% are obtained with models and tasks not used in the reward, showing the learned bit allocation transfers beyond the training objective. The Section 4.1 filter that excludes streams with zero x264 detection precision across most target bit-rates is a disclosed dataset-selection and external-validity caveat, not a circular step: on such streams the raw-frame detector provides an empty or constant reward, so the agent has no learning signal; the paper states this filter explicitly. No load-bearing self-citations or imported uniqueness theorems appear in the derivation. The Limitations section (Section 6) notes training-time and resolution-generalization constraints but does not hide the filter. Overall, the central claim is supported by held-out generalization and independent transfer evaluations, so the circularity score is 0.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central result rests on three free hyperparameters and three domain assumptions. The reward weighting lambda and the coarsening factor are chosen by hand and affect performance. The pseudo-ground-truth assumption and the dataset-filtering assumption are the most consequential for external validity.

free parameters (3)
  • lambda (reward weighting) = 20
    Chosen by hand in the reward r = r_bitrate + lambda * r_task; not tuned per task, but its scale depends on the relative magnitudes of the two rewards.
  • reward-prediction loss weight = 0.1
    Weight of the block-level reward-prediction MSE added to PPO loss (Appendix 7.7).
  • action-space coarsening factor = task-dependent optimum (Appendix 7.5)
    The action map is produced at lower resolution and upsampled; Figure 13 shows each task has a different optimal ratio, so this is a fitted-by-hand hyperparameter.
assumptions (3)
  • domain assumption The compression process is well-modeled as an MDP with frame-level time steps and Markovian state from encoder statistics.
    Used to justify PPO training in Section 3.1; if the encoder statistics do not capture relevant state, the policy cannot plan bit allocation.
  • domain assumption Raw-frame task model outputs are a reliable pseudo-ground-truth for evaluating reconstructed frames.
    Introduced in Section 3.1 as r_task; if the task model errs on raw frames, the reward is noisy and can misguide the agent.
  • domain assumption The filtered subset of BDD100K streams is representative of the target deployment distribution.
    Section 4.1 removes streams with zero precision across most bit-rates; the claim applies to this subset, not necessarily to all driving videos.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RL-RC-DoT: A Block-level RL agent for Task-Aware Video Compression." pith.science (2026). https://pith.science/paper/H3IQQS4V

@misc{pith2026250112216,
  author       = {Pith},
  title        = {Pith review of: RL-RC-DoT: A Block-level RL agent for Task-Aware Video Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H3IQQS4V}},
  note         = {Machine review of arXiv:2501.12216}
}
read the original abstract

Video encoders optimize compression for human perception by minimizing reconstruction error under bit-rate constraints. In many modern applications such as autonomous driving, an overwhelming majority of videos serve as input for AI systems performing tasks like object recognition or segmentation, rather than being watched by humans. It is therefore useful to optimize the encoder for a downstream task instead of for perceptual image quality. However, a major challenge is how to combine such downstream optimization with existing standard video encoders, which are highly efficient and popular. Here, we address this challenge by controlling the Quantization Parameters (QPs) at the macro-block level to optimize the downstream task. This granular control allows us to prioritize encoding for task-relevant regions within each frame. We formulate this optimization problem as a Reinforcement Learning (RL) task, where the agent learns to balance long-term implications of choosing QPs on both task performance and bit-rate constraints. Notably, our policy does not require the downstream task as an input during inference, making it suitable for streaming applications and edge devices such as vehicles. We demonstrate significant improvements in two tasks, car detection, and ROI (saliency) encoding. Our approach improves task performance for a given bit rate compared to traditional task agnostic encoding methods, paving the way for more efficient task-aware video compression.

Figures

Figures reproduced from arXiv: 2501.12216 by the authors.

Figure 1
Figure 1. RL-RC-DoT workflow. Our proposed solution to the block-level control for a downstream task. RL-RC-DoT takes encoder statistics as input and outputs a block-level delta QP map. We then evaluate the difference in downstream task performance between the reconstructed frame and the raw frame. The reward contains both global score as reward and block-level score. action space is very high-dimensional and we discuss be￾lo… view at source ↗
Figure 2
Figure 2. RD curves for Car detection task for differ￾ent streams (color). Com￾paring standard x264 (dashed lines) with RL-RC-DoT (solid lines). Curves show 3 exam￾ple streams, demonstrating how RL-RC-DoT improves quality across the range of bit-rate val￾ues. (a) Car detection precision (b) recall (c) PSNR. agent is trained using Stable-Baselines3 [30] implemented PPO with the reward function described in Section 3. We augmen… view at source ↗
Figure 3
Figure 3. Car detection example re￾sult. (a) detection output on x264 re￾constructed frame, (b) output on raw frame and (c) output on RL-RC-DoT reconstructed frame. Notice that both RL-RC-DoT and x264 used the same target bit-rate [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Example frame analysis. Red areas denote more bits allocate in the QP map ((a) x264 allocated bits,(c) RL-RC-DoT allocated bits ) and (b) Eigen-CAM values (middle). qualitative examples of our method’s performance, visually illustrating the enhanced quality in salient …
Figure 5
Figure 5. Figure 5: RD-curves for 3 videos (color) for ROI-encoding. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Saliency weighted PSNR re￾sults. (a) x264 reconstructed frame, (b) Saliency map of raw frame, ex￾tracted with [22] (c) RL-RC-DoT re￾constructed frame. Notice that both RL-RC-DoT and x264 used the same target bit-rate with a different task, we still observe an improved …
Figure 7
Figure 7. Figure 7: RL-RC-DoT agent architecture; Input is the statistics from the encoder, the output is the delta QP map [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Car detection example result. (a) detection output on x264 reconstructed frame, (b) output on raw frame and (c) output on [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Saliency weighted PSNR results. (a) x264 reconstructed frame, (b) Saliency map of raw frame, extracted with [ [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Car segmentation result comparison. (a) segmentation output on x264 reconstructed frame, (b) output on raw frame and (c) [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: Distribution of average KL divergence between QP maps and Eigen-CAM. [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: PSNR BD-rate to detection precision BD-rate, where each point represent a single stream in the test set [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: The effect of action space resolution on the BD-rate for both tasks [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 45 canonical work pages

  1. [19]

    Task-driven semantic cod- ing via reinforcement learning

    Xin Li, Jun Shi, and Zhibo Chen. Task-driven semantic cod- ing via reinforcement learning. IEEE Transactions on Image Processing, 30:6307–6320, 2021. 1, 2, 4, 5

    work page 2021

  2. [44]

    Hierarchical reinforcement learn- ing based video semantic coding for segmentation

    Guangqi Xie, Xin Li, Shiqi Lin, Zhibo Chen, Li Zhang, Kai Zhang, and Yue Li. Hierarchical reinforcement learn- ing based video semantic coding for segmentation. In 2022 IEEE International Conference on Visual Communications and Image Processing (VCIP), pages 1–5. IEEE, 2022. 1, 2, 4, 5

    work page 2022

  3. [1]

    Image and video compression standards: algorithms and architec- tures

    Vasudev Bhaskaran and Konstantinos Konstantinides. Image and video compression standards: algorithms and architec- tures. 1997. 1

    work page 1997

  4. [2]

    Calculation of average psnr differences between rd-curves

    Gisle Bjøntegaard. Calculation of average psnr differences between rd-curves. 2001. 5

    work page 2001

  5. [3]

    Overview of the versatile video coding (vvc) standard and its applica- tions

    Benjamin Bross, Ye-Kui Wang, Yan Ye, Shan Liu, Jianle Chen, Gary J Sullivan, and Jens-Rainer Ohm. Overview of the versatile video coding (vvc) standard and its applica- tions. IEEE Transactions on Circuits and Systems for Video Technology, 31(10):3736–3764, 2021. 1

    work page 2021

  6. [4]

    A novel video coding strategy in hevc for object detection

    Qi Cai, Zhifeng Chen, Dapeng Oliver Wu, Shan Liu, and Xiang Li. A novel video coding strategy in hevc for object detection. IEEE Transactions on Circuits and Systems for Video Technology, 31(12):4924–4937, 2021. 2

    work page 2021

  7. [5]

    Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolu- tion, and fully connected crfs

    Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L Yuille. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolu- tion, and fully connected crfs. IEEE transactions on pattern analysis and machine intelligence, 40(4):834–848, 2017. 7

    work page 2017

  8. [6]

    Re- inforcement learning for hevc/h

    Lian-Ching Chen, Jun-Hao Hu, and Wen-Hsiao Peng. Re- inforcement learning for hevc/h. 265 frame-level bit alloca- tion. In 2018 IEEE 23rd International Conference on Digital Signal Processing (DSP), pages 1–5. IEEE, 2018. 2

    work page 2018

Show all 50 references
  1. [7]

    Sieve: Semantically encoded video ana- lytics on edge and cloud

    Tarek Elgamal, Shu Shi, Varun Gupta, Rittwik Jana, and Klara Nahrstedt. Sieve: Semantically encoded video ana- lytics on edge and cloud. In 2020 IEEE 40th International Conference on Distributed Computing Systems (ICDCS), pages 1383–1388. IEEE, 2020. 2

    2020

  2. [8]

    Video coding for machines with feature- based rate-distortion optimization

    Kristian Fischer, Fabian Brand, Christian Herglotz, and Andr´e Kaup. Video coding for machines with feature- based rate-distortion optimization. In 2020 IEEE 22nd International Workshop on Multimedia Signal Processing (MMSP), pages 1–6. IEEE, 2020. 2, 5

    2020

  3. [9]

    Video compression for object detection al- gorithms

    Leonardo Galteri, Marco Bertini, Lorenzo Seidenari, and Al- berto Del Bimbo. Video compression for object detection al- gorithms. In 2018 24th International Conference on Pattern Recognition (ICPR), pages 3007–3012. IEEE, 2018. 2

    2018

  4. [10]

    Task-aware encoder control for deep video com- pression

    Xingtong Ge, Jixiang Luo, Xinjie Zhang, Tongda Xu, Guo Lu, Dailan He, Jing Geng, Yan Wang, Jun Zhang, and Hong- wei Qin. Task-aware encoder control for deep video com- pression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 26036– ...

    2024

  5. [11]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Gir- shick. Mask r-cnn. In Proceedings of the IEEE international conference on computer vision, pages 2961–2969, 2017. 5

    2017

  6. [12]

    A dual-critic reinforcement learning frame- work for frame-level bit allocation in hevc/h

    Yung-Han Ho, Guo-Lun Jin, Yun Liang, Wen-Hsiao Peng, and Xiaobo Li. A dual-critic reinforcement learning frame- work for frame-level bit allocation in hevc/h. 265. In 2021 Data compression conference (DCC), pages 13–22. IEEE,

    2021

  7. [13]

    Neural frank-wolfe policy optimiza- tion for region-of-interest intra-frame coding with hevc/h

    Yung-Han Ho, Chia-Hao Kao, Wen-Hsiao Peng, and Ping-Chun Hsieh. Neural frank-wolfe policy optimiza- tion for region-of-interest intra-frame coding with hevc/h

  8. [14]

    Re- inforcement learning for hevc/h

    Jun-Hao Hu, Wen-Hsiao Peng, and Chia-Hua Chung. Re- inforcement learning for hevc/h. 265 intra-frame rate con- trol. In 2018 IEEE International Symposium on Circuits and Systems (ISCAS), pages 1–5. IEEE, 2018. 2

    2018

  9. [15]

    ultralytics/yolov5

    Glenn Jocher. ultralytics/yolov5. https://github. com/ultralytics/ultralytics, 2020. 4, 5, 6

    2020

  10. [16]

    A new quality model for object detection using compressed videos

    Lingchao Kong, Rui Dai, and Yuchi Zhang. A new quality model for object detection using compressed videos. In2016 IEEE International Conference on Image Processing (ICIP), pages 3797–3801. IEEE, 2016. 3

    2016

  11. [17]

    Software and hardware hevc encoding

    Jan Kufa and Tomas Kratochvil. Software and hardware hevc encoding. In 2017 International Conference on Systems, Signals and Image Processing (IWSSIP), pages 1–5. IEEE,

    2017

  12. [18]

    Task-aware dis- tributed source coding under dynamic bandwidth

    Po-han Li, Sravan Kumar Ankireddy, Ruihan Philip Zhao, Hossein Nourkhiz Mahjoub, Ehsan Moradi Pari, Ufuk Topcu, Sandeep Chinchali, and Hyeji Kim. Task-aware dis- tributed source coding under dynamic bandwidth. Advances in Neural Information Processing Systems, 36, 2024. 2

    2024

  13. [20]

    Ssd: Single shot multibox detector

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Chris- tian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexan- der C Berg. Ssd: Single shot multibox detector. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 1...

    2016

  14. [21]

    Region-of- interest based resource allocation for conversational video communication of h

    Yang Liu, Zheng Guo Li, and Yeng Chai Soh. Region-of- interest based resource allocation for conversational video communication of h. 264/avc. IEEE transactions on circuits and systems for video technology, 18(1):134–139, 2008. 5

    2008

  15. [22]

    Transalnet: Towards perceptually relevant visual saliency prediction

    Jianxun Lou, Hanhe Lin, David Marshall, Dietmar Saupe, and Hantao Liu. Transalnet: Towards perceptually relevant visual saliency prediction. Neurocomputing, 2022. 5, 8, 6

    2022

  16. [23]

    Dvc: An end-to-end deep video compression framework

    Guo Lu, Wanli Ouyang, Dong Xu, Xiaoyun Zhang, Chun- lei Cai, and Zhiyong Gao. Dvc: An end-to-end deep video compression framework. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11006–11015, 2019. 1

    2019

  17. [24]

    Muzero with self- competition for rate control in vp9 video compression.arXiv preprint arXiv:2202.06626, 2022

    Amol Mandhane, Anton Zhernov, Maribeth Rauh, Chenjie Gu, Miaosen Wang, Flora Xue, Wendy Shang, Derek Pang, Rene Claus, Ching-Han Chiang, et al. Muzero with self- competition for rate control in vp9 video compression.arXiv preprint arXiv:2202.06626, 2022. 2

    2022 arXiv

  18. [25]

    Neural rate control for video encoding using imitation learning

    Hongzi Mao, Chenjie Gu, Miaosen Wang, Angie Chen, Nevena Lazic, Nir Levine, Derek Pang, Rene Claus, Maris- abel Hechtman, Ching-Han Chiang, et al. Neural rate control for video encoding using imitation learning. arXiv preprint arXiv:2012.05339, 2020. 2

    2012 arXiv

  19. [26]

    x264: A high per- formance h

    Loren Merritt and Rahul Vanam. x264: A high per- formance h. 264/avc encoder. online] http://neuron2. net/library/avc/overview x264 v8 5. pdf, 2006. 1, 2, 5, 11

    2006

  20. [27]

    Eigen-cam: Class activation map using principal compo- nents

    Mohammed Bany Muhammad and Mohammed Yeasin. Eigen-cam: Class activation map using principal compo- nents. In 2020 international joint conference on neural networks (IJCNN), pages 1–7. IEEE, 2020. 6, 7

    2020

  21. [28]

    Factored action spaces in deep re- inforcement learning

    Thomas PIERROT, Valentin Mac ´e, Jean-Baptiste Sevestre, Louis Monier, Alexandre Laterre, Nicolas Perrin, Karim Be- guir, and Olivier Sigaud. Factored action spaces in deep re- inforcement learning. 2021. 3

    2021

  22. [29]

    The 2017 davis challenge on video object segmentation.arXiv preprint arXiv:1704.00675, 2017

    Jordi Pont-Tuset, Federico Perazzi, Sergi Caelles, Pablo Ar- bel´aez, Alex Sorkine-Hornung, and Luc Van Gool. The 2017 davis challenge on video object segmentation.arXiv preprint arXiv:1704.00675, 2017. 5

    2017 arXiv

  23. [30]

    Stable- baselines3: Reliable reinforcement learning implementa- tions

    Antonin Raffin, Ashley Hill, Adam Gleave, Anssi Kan- ervisto, Maximilian Ernestus, and Noah Dormann. Stable- baselines3: Reliable reinforcement learning implementa- tions. Journal of Machine Learning Research, 22(268):1–8,

  24. [31]

    Proximal policy optimization algo- rithms

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Rad- ford, and Oleg Klimov. Proximal policy optimization algo- rithms. arXiv preprint arXiv:1707.06347, 2017. 3, 2

    2017 arXiv

  25. [32]

    Reinforced bit allocation un- der task-driven semantic distortion metrics

    Jun Shi and Zhibo Chen. Reinforced bit allocation un- der task-driven semantic distortion metrics. In 2020 IEEE international symposium on circuits and systems (ISCAS), pages 1–5. IEEE, 2020. 1, 5, 3

    2020

  26. [33]

    The need for medically aware video compression in gastroenterology.arXiv preprint arXiv:2211.01472, 2022

    Joel Shor and Nick Johnston. The need for medically aware video compression in gastroenterology.arXiv preprint arXiv:2211.01472, 2022. 2

    2022 arXiv

  27. [34]

    Video- analytics task-aware quad-tree partitioning and quantization for hevc

    Praneet Singh, Edward J Delp, and Amy R Reibman. Video- analytics task-aware quad-tree partitioning and quantization for hevc. In 2022 IEEE International Conference on Image Processing (ICIP), pages 2936–2940. IEEE, 2022. 2

    2022

  28. [35]

    Overview of the high efficiency video coding (hevc) standard

    Gary J Sullivan, Jens-Rainer Ohm, Woo-Jin Han, and Thomas Wiegand. Overview of the high efficiency video coding (hevc) standard. IEEE Transactions on circuits and systems for video technology, 22(12):1649–1668, 2012. 1

    2012

  29. [36]

    Sutton and Andrew G

    Richard S. Sutton and Andrew G. Barto. Reinforcement Learning: An Introduction. The MIT Press, Cambridge, MA,

  30. [37]

    Converting video formats with ffmpeg

    Suramya Tomar. Converting video formats with ffmpeg. Linux Journal, 2006(146):10, 2006. 11

    2006

  31. [38]

    Reinforce- ment learning in continuous action spaces

    Hado Van Hasselt and Marco A Wiering. Reinforce- ment learning in continuous action spaces. In 2007 IEEE International Symposium on Approximate Dynamic Programming and Reinforcement Learning, pages 272–279. IEEE, 2007. 3

    2007

  32. [39]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13(4):600–612, 2004. 3

    2004

  33. [40]

    Stephan Wenger. H. 264/avc over ip. IEEE transactions on circuits and systems for video technology, 13(7):645–656,

  34. [41]

    Overview of the h

    Thomas Wiegand, Gary J Sullivan, Gisle Bjontegaard, and Ajay Luthra. Overview of the h. 264/avc video coding stan- dard. IEEE Transactions on circuits and systems for video technology, 13(7):560–576, 2003. 5, 3

    2003

  35. [42]

    On annotation-free optimization of video coding for machines

    Marc Windsheimer, Fabian Brand, and Andr ´e Kaup. On annotation-free optimization of video coding for machines. arXiv preprint arXiv:2406.07938, 2024. 2

    2024 arXiv

  36. [43]

    Qs- nerv: Real-time quality-scalable decoding with neural repre- sentation for videos

    Chang Wu, Guancheng Quan, Gang He, Xin-Quan Lai, Yun- song Li, Wenxin Yu, Xianmeng Lin, and Cheng Yang. Qs- nerv: Real-time quality-scalable decoding with neural repre- sentation for videos. In ACM Multimedia 2024, 2024. 2

    2024

  37. [45]

    Mastering atari games with limited data

    Weirui Ye, Shaohuai Liu, Thanard Kurutach, Pieter Abbeel, and Yang Gao. Mastering atari games with limited data. Advances in neural information processing systems, 34: 25476–25488, 2021. 4

    2021

  38. [46]

    Bdd100k: A diverse driving dataset for heteroge- neous multitask learning

    Fisher Yu, Haofeng Chen, Xin Wang, Wenqi Xian, Yingy- ing Chen, Fangchen Liu, Vashisht Madhavan, and Trevor Darrell. Bdd100k: A diverse driving dataset for heteroge- neous multitask learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition,...

    2020

  39. [47]

    Competitive learning for achieving content-specific filters in video coding for machines

    Honglei Zhang, Jukka I Ahonen, Nam Le, Ruiying Yang, and Francesco Cricri. Competitive learning for achieving content-specific filters in video coding for machines. arXiv preprint arXiv:2406.12367, 2024. 2

    2024 arXiv

  40. [48]

    Bytetrack: Multi-object tracking by associating every detection box

    Yifu Zhang, Peize Sun, Yi Jiang, Dongdong Yu, Fucheng Weng, Zehuan Yuan, Ping Luo, Wenyu Liu, and Xinggang Wang. Bytetrack: Multi-object tracking by associating every detection box. In European conference on computer vision, pages 1–21. Springer, 2022. 5

    2022

  41. [50]

    Environment and agent details 7.1.1

    Appendix 7.1. Environment and agent details 7.1.1. Environment details Global encoder statistics used as state information Next frame x264 selected QP value Next frame number Current bitstream size Current frame x264 selected QP value Average QP Percentages of I type Macro Blo...

  42. [265]

    IEEE, 2022

    In 2022 IEEE International Conference on Visual Communications and Image Processing (VCIP), pages 1–5. IEEE, 2022. 2

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.