Cost-Effective Active Learning (CEAL) for Deep Image Classification Implementation with keras
Model - Resnet18v2
Dataset - Cifar10
python CEAL_keras.py
-h, --help show this help message and exit
-verbose VERBOSE Verbosity mode. 0 = silent, 1 = progress bar, 2 = one
line per epoch. default: 0
-epochs EPOCHS Number of epoch to train. default: 5
-batch_size BATCH_SIZE
Number of samples per gradient update. default: 32
-chkt_filename CHKT_FILENAME
Model Checkpoint filename to save
-t FINE_TUNNING_INTERVAL, --fine_tunning_interval FINE_TUNNING_INTERVAL
Fine-tuning interval. default: 1
-T MAXIMUM_ITERATIONS, --maximum_iterations MAXIMUM_ITERATIONS
Maximum iteration number. default: 10
-i INITIAL_ANNOTATED_PERC, --initial_annotated_perc INITIAL_ANNOTATED_PERC
Initial Annotated Samples Percentage. default: 0.1
-dr THRESHOLD_DECAY, --threshold_decay THRESHOLD_DECAY
Threshold decay rate. default: 0.0033
-delta DELTA High confidence samples selection threshold. default:
0.05
-K UNCERTAIN_SAMPLES_SIZE, --uncertain_samples_size UNCERTAIN_SAMPLES_SIZE
Uncertain samples selection size. default: 2000
-uc UNCERTAIN_CRITERIA, --uncertain_criteria UNCERTAIN_CRITERIA
Uncertain selection Criteria: 'lc' (Least Confidence),
'ms' (Margin Sampling), 'en' (Entropy). default: lc
-ce COST_EFFECTIVE, --cost_effective COST_EFFECTIVE
whether to use Cost Effective high confidence sample
pseudo-labeling. default: True