Skip to content

Latest commit

 

History

History
39 lines (29 loc) · 3.1 KB

MODEL_ZOO.md

File metadata and controls

39 lines (29 loc) · 3.1 KB
Raw

I3D, I3D Nonlocal

Use following model configs.

  • model_configs/i3d_resnet50.py
  • model_configs/i3dnonlocal_resnet50.py

Kinetics

Reference: facebookresearch/SlowFast

architecture depth pretrain frame length x sample rate top1 top5 model config
I3D R50 - 8 x 8 73.5 90.8 link Kinetics/c2/I3D_8x8_R50
I3D NLN R50 - 8 x 8 74.0 91.1 link Kinetics/c2/I3D_NLN_8x8_R50

Reference: facebookresearch/video-nonlocal-net

script input frames freeze bn? 3D conv? non-local? top1 in paper top5 model logs
run_i3d_baseline_400k_32f.sh 32 - Yes - 73.6 73.3 90.8 link link
run_i3d_nlnet_400k_32f.sh 32 - Yes Yes 74.9 74.9 91.6 link link

HMDB-51

This model is trained with PyVideoAI.
Top1/5 accuracy is calculated using 1 spatial centre crop and 5 temporal crops.
dataset_configs/hmdb.py
model_configs/i3d_resnet50.py
exp_configs/hmdb/i3d_resnet50-crop224_lr0001_batch8_8x8_largejit_plateau_1scrop5tcrop_split1.py

architecture Pretrain frame length x sampling stride Top1 (highest/last) Top5 (highest/last) config, log model (last) TensorBoard
I3D-ResNet50 Kinetics 8 x 8 73.20 / 72.94 94.05 / 94.05 link link link

ImageNet

Reference: facebookresearch/SlowFast

architecture depth Top1 Top5 model
ResNet R50 23.6 6.8 link