SNeRV

September 12, 2024 ยท View on GitHub

This is the repository of SNeRV: Spectra-preserving Neural Representation for Video - ECCV 2024

Get started

  • Prepare environment
pip install -r requirements.txt 
  • Prepare dataset files at data/

Examples

SNeRV

CUDA_VISIBLE_DEVICES=0 python train_snerv.py  --outf exp_1 --data_path  data/bunny --vid bunny \
    --model snerv \
    --enc_strds 5 4 2 2 2 --dec_strds 5 4 2 2 2 \
    --num_blocks 6 --fc_dim 111 \
    --crop_list 640_1280 --data_split 1_1_1 --eval_freq 50 -e 300 \
    --lr 0.001 --suffix SNeRV --emb_size 0

Temporal extension of SNeRV

CUDA_VISIBLE_DEVICES=0 python train_snerv_t.py  --outf exp_1 --data_path  data/bunny --vid bunny \
    --model snerv_t \
    --enc_strds 5 4 2 2 2 --dec_strds 5 4 2 2 2 \
    --num_blocks 6 --fc_dim 103 \
    --crop_list 640_1280 --data_split 1_1_1 --eval_freq 50 -e 300 \
    --lr 0.001 --suffix SNeRV_T --emb_size 20
  • We also provide SNeRV_T(2D) model as --model snerv_t_2d

Acknowledgement

The implementation is based on HNeRV.

Citation