Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
PNVC: Towards Practical INR-based Video Compression
Authors: Ge Gao, Ho Man Kwan, Fan Zhang, David Bull
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The proposed PNVC is pre-trained on Vimeo90k (Xue et al. 2019) and evaluated on UVG (Mercat, Viitanen, and Vanne 2020), MCL-JCV (Wang et al. 2016), and HEVC B-E (Bossen et al. 2013), where the results for HEVC sequences are provided in the Supplementary. ... The BD-rate results in Table 1 show that PNVC (both LD and RA) outperforms state-of-the-art INR-based codecs and proves competitive with leading conventional and neural video codecs. ... To validate the contribution of each design component in PNVC, we performed ablations with various model variants. |
| Researcher Affiliation | Academia | Visual Information Laboratory, University of Bristol Bristol, BS1 5DD, UK EMAIL |
| Pseudocode | No | The paper describes the proposed methods and architectures through figures (e.g., Figure 2, Figure 3) and mathematical equations, but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code https://github.com/ge1-gao/PNVC |
| Open Datasets | Yes | The proposed PNVC is pre-trained on Vimeo90k (Xue et al. 2019) and evaluated on UVG (Mercat, Viitanen, and Vanne 2020), MCL-JCV (Wang et al. 2016), and HEVC B-E (Bossen et al. 2013), where the results for HEVC sequences are provided in the Supplementary. |
| Dataset Splits | No | The paper describes 'Test conditions' for video coding configurations like Low Delay (LD) and Random Access (RA), defining GOP lengths and frame types for evaluation. However, it does not provide specific train/validation/test splits for the datasets themselves (e.g., Vimeo90k, UVG, MCL-JCV). |
| Hardware Specification | Yes | These are measured by averaging over a GOP of 1080P videos, including a full roll-out of the entropy coding process, on a PC with an NVIDIA 3090 GPU and an Intel Core i712700 CPU. |
| Software Dependencies | No | The paper mentions various models and techniques used (e.g., ELIC, Spy Net, Conv LSTM), but it does not specify versions for any programming languages, libraries, or frameworks (e.g., Python, PyTorch, TensorFlow, CUDA). |
| Experiment Setup | No | The paper details the optimization strategy, including pretraining then overfitting, loss functions (rate-distortion loss, MSE, mixed distortion loss), and distortion metrics (PSNR, MS-SSIM). However, it does not specify concrete hyperparameters such as learning rates, batch sizes, number of epochs, or the specific optimizers and their configurations. |