imdb-wiki-dir

IMDB-WIKI-DIR

Installation

Prerequisites

1. Download and extract IMDB faces and WIKI faces respectively using

python download_imdb_wiki.py

2. (Optional) We have provided required IMDB-WIKI-DIR meta file imdb_wiki.csv to set up balanced val/test set in folder ./data. To reproduce the results in the paper, please directly use this file. You can also generate it using

python data/create_imdb_wiki.py
python data/preprocess_imdb_wiki.py

Dependencies

• PyTorch (>= 1.2, tested on 1.6)
• tensorboard_logger
• numpy, pandas, scipy, tqdm, matplotlib, PIL, wget

Getting Started

Stage 1. Train the base model

Train a vanilla model as the base model:

python train.py

Alternatively, a pretrained base model can be downloaded from here.

Stage 2. Train a model using RRT + Balanced MSE

GAI

# preprocess gmm
python preprocess_gmm.py

python train.py \
--balanced_metric \
--bmse --imp gai --init_noise_sigma 10. --gmm gmm.pkl \
--sigma_lr 0.01 \
--retrain_fc --pretrained <path_to_base_model_ckpt> \
--lr 0.0001 --epoch 10

BMC

python train.py \
--balanced_metric \
--bmse --imp bmc --init_noise_sigma 10. \
--retrain_fc --pretrained <path_to_base_model_ckpt> \
--lr 0.0001 --epoch 10

BNI

python train.py \
--balanced_metric \
--bmse --imp bni --init_noise_sigma 10. \
--sigma_lr 0.01 \
--retrain_fc --pretrained <path_to_base_model_ckpt>  \
--lr 0.0001 --epoch 10

Evaluate a trained checkpoint

python train.py [...evaluation model arguments...] --evaluate --resume <path_to_evaluation_ckpt>

Reproduced Benchmarks and Model Zoo

We provide below reproduced results on IMDB-WIKI-DIR (base method Vanilla, metric MAE).

Model	Overall	Many-Shot	Medium-Shot	Few-Shot	Download
Vanilla	13.923	7.323	15.925	32.778	model
GAI	12.690	7.589	12.880	28.307	model
BMC	12.654	7.649	12.689	28.097	model
BNI	12.650	7.647	12.696	28.077	model