miracleyoo / 6-pic-vote-mobilenet

A directory which aims to do card classification based on 6 base pictures generated from the original image.

6-pic-vote-mobilenet

Config

USE_CUDA : Whether to use GPU.

LOAD_SAVED_MOD : Whether to load saved model.

SAVE_TEMP_MODEL : Whether to save temporary model while training.

SAVE_BEST_MODEL : Whether to save best model while training.

BEST_MODEL_BY_LOSS : Evaluate whether a model is the optimal one by loss or accuracy.

PRINT_BAD_CASE : Whether to print the bad case while predicting.

RUNNING_ON_JUPYTER : Whether the program is running on a Jupyter Notebook.

START_VOTE_PREDICT : Whether to start vote predicting or training.

START_PREDICT : Whether to start predicting or training.

TRAIN_ALL : Whether to train in all of the data (train_set and val_set).

TEST_ALL : Whether to validate all of the data (train_set and val_set).

TO_MULTI : Whether to use multiple GPU, if available.

ADD_SUMMARY : Whether to add net graph into tensorboard summary.

SAVE_PER_EPOCH : Save your temp model every n epoch.

BATCH_SIZE : Batch size of training.

VAL_BATCH_SIZE : Batch size of validating.

TENSOR_SHAPE : Tensor shape of your input (batch dim is not included).

DATALOADER_TYPE : Dataloader type of your data (only ImageFolder, SamplePairing, SixBatch)

OPTIMIZER : Optimizer type. It is a string which is not case sensitive.Currently Adam and SGD are supported. Add new optimizer in the ./models/BasicModule.py -> get_optimizer()

SGD_MOMENTUM : The momentum if SDG is chosen as optimizer.

TRAIN_DATA_RATIO : The Train_Val data ratio.

NUM_EPOCHS : The epochs you want to train your model.

NUM_CLASSES : The number of your input data's class.

NUM_VAL : The number of your validation data.

NUM_TRAIN : The number of your train data.

TOP_NUM : If top n accuracy is ok for your result, put the n here.

NUM_WORKERS : Number of workers used in the DataLoader.

CRITERION : The Loss Class used in your training process, which is an instance of a Loss Class.

LEARNING_RATE : Learning rate used in your optimizer.

TOP_VOTER : Top n votes in the 6 picture generated will count for the final result.

NET_SAVE_PATH : Where to save your trained model.

TRAIN_PATH : Where your training set is located.

VAL_PATH : Where your validating set is located.

CLASSES_PATH : Where to save your classes' name.

MODEL_NAME : The name of your model.

PROCESS_ID : The ID of the current training process, which is the marker of the trained models. Please change it when some config or crucial code is altered!

SUMMARY_PATH : Where to save your tensorboard summary.