def main(DATASET,
LABELS,
CLASS_IDS,
BATCH_SIZE,
ANNOTATION_FILE,
SEQ_SIZE=16,
STEP=16,
nstrokes=-1,
N_EPOCHS=25):
'''
Extract sequence features from AutoEncoder.
Parameters:
-----------
DATASET : str
path to the video dataset
LABELS : str
path containing stroke labels
CLASS_IDS : str
path to txt file defining classes, similar to THUMOS
BATCH_SIZE : int
size for batch of clips
SEQ_SIZE : int
no. of frames in a clip (min. 16 for 3D CNN extraction)
STEP : int
stride for next example. If SEQ_SIZE=16, STEP=8, use frames (0, 15), (8, 23) ...
partition : str
'all' / 'train' / 'test' / 'val' : Videos to be considered
nstrokes : int
partial extraction of features (do not execute for entire dataset)
Returns:
--------
trajectories, stroke_names
'''
###########################################################################
attn_utils.seed_everything(1234)
if not os.path.isdir(log_path):
os.makedirs(log_path)
# Read the strokes
# Divide the highlight dataset files into training, validation and test sets
train_lst, val_lst, test_lst = autoenc_utils.split_dataset_files(DATASET)
print("No. of training videos : {}".format(len(train_lst)))
# extract_of_features(feat_path, DATASET, LABELS, train_lst, val_lst)
features, stroke_names_id = attn_utils.read_feats(feat_path, feat, snames)
# get matrix of features from dictionary (N, vec_size)
vecs = []
for key in sorted(list(features.keys())):
vecs.append(features[key])
vecs = np.vstack(vecs)
vecs[np.isnan(vecs)] = 0
vecs[np.isinf(vecs)] = 0
#fc7 layer output size (4096)
INP_VEC_SIZE = vecs.shape[-1]
print("INP_VEC_SIZE = ", INP_VEC_SIZE)
km_filepath = os.path.join(log_path, km_filename)
# # Uncomment only while training.
if not os.path.isfile(km_filepath + "_C" + str(cluster_size) + ".pkl"):
km_model = make_codebook(vecs, cluster_size) #, model_type='gmm')
## # Save to disk, if training is performed
print("Writing the KMeans models to disk...")
pickle.dump(
km_model,
open(km_filepath + "_C" + str(cluster_size) + ".pkl", "wb"))
else:
# Load from disk, for validation and test sets.
km_model = pickle.load(
open(km_filepath + "_C" + str(cluster_size) + ".pkl", 'rb'))
print("Create numpy one hot representation for train features...")
onehot_feats = create_bovw_SA(features, stroke_names_id, km_model)
ft_path = os.path.join(log_path, "C" + str(cluster_size) + "_train.pkl")
with open(ft_path, "wb") as fp:
pickle.dump(onehot_feats, fp)
###########################################################################
features_val, stroke_names_id_val = attn_utils.read_feats(
feat_path, feat_val, snames_val)
print("Create numpy one hot representation for val features...")
onehot_feats_val = create_bovw_SA(features_val, stroke_names_id_val,
km_model)
ft_path_val = os.path.join(log_path, "C" + str(cluster_size) + "_val.pkl")
with open(ft_path_val, "wb") as fp:
pickle.dump(onehot_feats_val, fp)
###########################################################################
features_test, stroke_names_id_test = attn_utils.read_feats(
feat_path, feat_test, snames_test)
print("Create numpy one hot representation for test features...")
onehot_feats_test = create_bovw_SA(features_test, stroke_names_id_test,
km_model)
ft_path_test = os.path.join(log_path,
"C" + str(cluster_size) + "_test.pkl")
with open(ft_path_test, "wb") as fp:
pickle.dump(onehot_feats_test, fp)
###########################################################################
# Create a Dataset
train_dataset = StrokeFeaturePairsDataset(ft_path,
train_lst,
DATASET,
LABELS,
CLASS_IDS,
frames_per_clip=SEQ_SIZE,
extracted_frames_per_clip=2,
step_between_clips=STEP,
train=True)
val_dataset = StrokeFeaturePairsDataset(ft_path_val,
val_lst,
DATASET,
LABELS,
CLASS_IDS,
frames_per_clip=SEQ_SIZE,
extracted_frames_per_clip=2,
step_between_clips=STEP,
train=False)
# get labels
labs_keys, labs_values = attn_utils.get_cluster_labels(ANNOTATION_FILE)
# # created weighted Sampler for class imbalance
# samples_weight = attn_utils.get_sample_weights(train_dataset, labs_keys, labs_values,
# train_lst)
# sampler = WeightedRandomSampler(samples_weight, len(samples_weight))
train_loader = DataLoader(dataset=train_dataset,
batch_size=BATCH_SIZE,
shuffle=True)
# sampler=sampler, worker_init_fn=np.random.seed(12))
val_loader = DataLoader(dataset=val_dataset,
batch_size=BATCH_SIZE,
shuffle=False)
data_loaders = {"train": train_loader, "test": val_loader}
num_classes = len(list(set(labs_values)))
###########################################################################
# load model and set loss function
ntokens = cluster_size # the size of vocabulary
emsize = 200 # embedding dimension
nhid = 200 # the dimension of the feedforward network model in nn.TransformerEncoder
nlayers = 2 # the number of nn.TransformerEncoderLayer in nn.TransformerEncoder
nhead = 2 # the number of heads in the multiheadattention models
dropout = 0.2 # the dropout value
model = tt.TransformerModelSA(ntokens, emsize, nhead, nhid, nlayers,
dropout).to(device)
# Setup the loss fxn
# criterion = nn.CrossEntropyLoss()
criterion = nn.MSELoss()
model = model.to(device)
# print("Params to learn:")
params_to_update = []
for name, param in model.named_parameters():
if param.requires_grad == True:
params_to_update.append(param)
print("\t", name)
# # Observe that all parameters are being optimized
## optimizer_ft = torch.optim.Adam(model.parameters(), lr=0.001)
# optimizer = torch.optim.SGD(model.parameters(), lr=0.001, momentum=0.9)
#
# # Decay LR by a factor of 0.1 every 7 epochs
# scheduler = StepLR(optimizer, step_size=10, gamma=0.1)
lr = 5.0 # learning rate
optimizer = torch.optim.SGD(model.parameters(), lr=lr)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, 2.0, gamma=0.95)
###########################################################################
# Training the model
# start = time.time()
#
# model = train_model(features, stroke_names_id, model, data_loaders, criterion,
# optimizer, scheduler, labs_keys, labs_values,
# num_epochs=N_EPOCHS)
#
# end = time.time()
#
## # save the best performing model
# save_model_checkpoint(log_path, model, N_EPOCHS,
# "S"+str(SEQ_SIZE)+"C"+str(cluster_size)+"_SGD")
# Load model checkpoints
model = load_weights(
log_path, model, N_EPOCHS,
"S" + str(SEQ_SIZE) + "C" + str(cluster_size) + "_SGD")
# print("Total Execution time for {} epoch : {}".format(N_EPOCHS, (end-start)))
###########################################################################
###########################################################################
# Extract attention model features
if not os.path.isfile(os.path.join(log_path, "trans_feats.pkl")):
if not os.path.exists(log_path):
os.makedirs(log_path)
# # Extract Grid OF / HOOF features {mth = 2, and vary nbins}
print("Training extraction ... ")
feats_dict, stroke_names = extract_trans_feats(model,
DATASET,
LABELS,
CLASS_IDS,
BATCH_SIZE,
SEQ_SIZE,
SEQ_SIZE - 1,
partition='train',
nstrokes=nstrokes,
base_name=log_path)
with open(os.path.join(log_path, "trans_feats.pkl"), "wb") as fp:
pickle.dump(feats_dict, fp)
with open(os.path.join(log_path, "trans_snames.pkl"), "wb") as fp:
pickle.dump(stroke_names, fp)
if not os.path.isfile(os.path.join(log_path, "trans_feats_val.pkl")):
print("Validation extraction ....")
feats_dict_val, stroke_names_val = extract_trans_feats(
model,
DATASET,
LABELS,
CLASS_IDS,
BATCH_SIZE,
SEQ_SIZE,
SEQ_SIZE - 1,
partition='val',
nstrokes=nstrokes,
base_name=log_path)
with open(os.path.join(log_path, "trans_feats_val.pkl"), "wb") as fp:
pickle.dump(feats_dict_val, fp)
with open(os.path.join(log_path, "trans_snames_val.pkl"), "wb") as fp:
pickle.dump(stroke_names_val, fp)
if not os.path.isfile(os.path.join(log_path, "trans_feats_test.pkl")):
print("Testing extraction ....")
feats_dict_val, stroke_names_val = extract_trans_feats(
model,
DATASET,
LABELS,
CLASS_IDS,
BATCH_SIZE,
SEQ_SIZE,
SEQ_SIZE - 1,
partition='test',
nstrokes=nstrokes,
base_name=log_path)
with open(os.path.join(log_path, "trans_feats_test.pkl"), "wb") as fp:
pickle.dump(feats_dict_val, fp)
with open(os.path.join(log_path, "trans_snames_test.pkl"), "wb") as fp:
pickle.dump(stroke_names_val, fp)
# call count_paramters(model) for displaying total no. of parameters
print("#Parameters : {} ".format(autoenc_utils.count_parameters(model)))
return 0
def main(base_name, nbins=10, ft_type='2dcnn', cluster_size=10):
"""
Function to extract orientation features and find the directions of strokes,
using LDA model/clustering and evaluate on three cluster analysis on highlights.
The videos can be visualized by writing trimmed class videos into their respective
classes.
Parameters:
------
base_name: path to the wts, losses, predictions and log files
use_gpu: True if training to be done on GPU, False for CPU
"""
seed = 1234
print(60 * "#")
#####################################################################
# Form dataloaders
# train_lst_main_ext = get_main_dataset_files(MAIN_DATASET) #with extensions
# train_lst_main = [t.rsplit('.', 1)[0] for t in train_lst_main_ext] # remove the extension
# val_lst_main_ext = get_main_dataset_files(VAL_DATASET)
# val_lst_main = [t.rsplit('.', 1)[0] for t in val_lst_main_ext]
# Divide the samples files into training set, validation and test sets
train_lst, val_lst, test_lst = utils.split_dataset_files(DATASET)
#print("c3dWinSize : {}".format(c3dWinSize))
# form the names of the list of label files, should be at destination
train_lab = [f + ".json" for f in train_lst]
val_lab = [f + ".json" for f in val_lst]
test_lab = [f + ".json" for f in test_lst]
# train_lab_main = [f+".json" for f in train_lst_main]
# val_lab_main = [f+".json" for f in val_lst_main]
# get complete path lists of label files
tr_labs = [os.path.join(LABELS, f) for f in train_lab]
val_labs = [os.path.join(LABELS, f) for f in val_lab]
# tr_labs_main = [os.path.join(MAIN_LABELS, f) for f in train_lab_main]
# val_labs_main = [os.path.join(VAL_LABELS, f) for f in val_lab_main]
#####################################################################
sizes = [
utils.getNFrames(os.path.join(DATASET, f + ".avi")) for f in train_lst
]
val_sizes = [
utils.getNFrames(os.path.join(DATASET, f + ".avi")) for f in val_lst
]
# sizes_main = [utils.getNFrames(os.path.join(MAIN_DATASET, f)) for f in train_lst_main_ext]
# val_sizes_main = [utils.getNFrames(os.path.join(VAL_DATASET, f)) for f in val_lst_main_ext]
###########################################################################
# Merge the training highlights and main dataset variables
# train_lab.extend(train_lab_main)
# tr_labs.extend(tr_labs_main)
# sizes.extend(sizes_main)
print("No. of training videos : {}".format(len(train_lst)))
print("Size : {}".format(sizes))
# hlDataset = VideoDataset(tr_labs, sizes, seq_size=SEQ_SIZE, is_train_set = True)
# print(hlDataset.__len__())
#####################################################################
# Feature Extraction : (GRID OF / HOOF / 2D CNN / 3DCNN / IDT)
print("Feature Type : {} :: nClusters : {} ".format(ft_type, cluster_size))
#####################################################################
# read into dictionary {vidname: np array, ...}
BATCH_SIZE, SEQ_SIZE, STEP = 16, 16, 1
print("Loading features from disk...")
if not os.path.exists(base_name):
os.makedirs(base_name)
features, strokes_name_id = extract_feats(DATASET,
LABELS,
CLASS_IDS,
BATCH_SIZE,
SEQ_SIZE,
STEP,
extractor=ft_type,
model_path=model_path,
nclasses=NUM_TOPICS,
part='train')
with open(os.path.join(base_name, ft_type + "_feats_train.pkl"),
"wb") as fp:
pickle.dump(features, fp)
with open(os.path.join(base_name, ft_type + "_snames_train.pkl"),
"wb") as fp:
pickle.dump(strokes_name_id, fp)
with open(os.path.join(base_name, ft_type + "_feats_train.pkl"),
"rb") as fp:
features = pickle.load(fp)
with open(os.path.join(base_name, ft_type + "_snames_train.pkl"),
"rb") as fp:
strokes_name_id = pickle.load(fp)
#####################################################################
# get matrix of features from dictionary (N, vec_size)
vecs = []
for key in sorted(list(features.keys())):
vecs.append(features[key])
vecs = np.vstack(vecs)
vecs[np.isnan(vecs)] = 0
vecs[np.isinf(vecs)] = 0
#fc7 layer output size (4096)
INP_VEC_SIZE = vecs.shape[-1]
print("INP_VEC_SIZE = ", INP_VEC_SIZE)
km_filepath = os.path.join(base_name, km_filename)
# # Uncomment only while training.
if not os.path.isfile(km_filepath + "_C" + str(cluster_size) + ".pkl"):
km_model = make_codebook(vecs, cluster_size) #, model_type='gmm')
## # Save to disk, if training is performed
print("Writing the KMeans models to disk...")
pickle.dump(
km_model,
open(km_filepath + "_C" + str(cluster_size) + ".pkl", "wb"))
else:
# Load from disk, for validation and test sets.
km_model = pickle.load(
open(km_filepath + "_C" + str(cluster_size) + ".pkl", 'rb'))
###########################################################################
# Form the training dataset for supervised classification
# Assign the words (flow frames) to their closest cluster centres and count the
# frequency for each document(video). Create IDF bow dataframe by weighting
# df_train is (nVids, 50) for magnitude, with index as videonames
# print("Create a dataframe for C3D FC7 features...")
# df_train_c3d, words_train = create_bovw_c3d_traindf(features, \
# strokes_name_id, km_model, c3dWinSize)
print("Create a dataframe for HOOF features...")
df_train, words_train = create_bovw_df(features, strokes_name_id, km_model,\
base_name, "train")
# read the stroke annotation labels from text file.
vids_list = list(df_train.index)
labs_keys, labs_values = get_cluster_labels(ANNOTATION_FILE)
if min(labs_values) == 1:
labs_values = [l - 1 for l in labs_values]
labs_keys = [k.replace('.avi', '') for k in labs_keys]
train_labels = np.array(
[labs_values[labs_keys.index(v)] for v in vids_list])
###########################################################################
# apply_clustering(df_train, DATASET, LABELS, ANNOTATION_FILE, base_name)
###########################################################################
# print("Training stroke labels : ")
# print(train_labels)
# print(train_labels.shape)
# concat dataframe to contain features and corresponding labels
#df_train = pd.concat([df_train_mag, labs_df], axis=1)
###########################################################################
# Train SVM
clf = LinearSVC(verbose=False, random_state=124, max_iter=3000)
clf.fit(df_train, train_labels)
print("Training Complete.")
###########################################################################
# # Train a classifier on the features.
# print("LDA execution !!! ")
# #Run LDA
#
# # Get list of lists. Each sublist contains video cluster strIDs (words).
# # Eg. [["39","29","39","39","0", ...], ...]
# doc_clean = [doc.split() for doc in words_train]
# #print(doc_clean)
# diction=corpora.Dictionary(doc_clean) # Form a dictionary
# print("printing dictionary after corp {} ".format(diction))
# doc_term_matrix = [diction.doc2bow(doc) for doc in doc_clean]
# #dictionary = corpora.Dictionary(diction)
#
# # Inference using the data.
# ldamodel_obj = gensim.models.ldamodel.LdaModel(doc_term_matrix, \
# num_topics = NUM_TOPICS, id2word=diction, passes=10, \
# random_state=seed)
## ldamodel_obj = gensim.models.ldaseqmodel.LdaSeqModel(doc_term_matrix, \
## num_topics=3, time_slice=[351])
## ldamodel_obj = gensim.models.LsiModel(doc_term_matrix, num_topics=3, \
## id2word = diction)
#
# print("training complete saving to disk ")
# #save model to disk
# joblib.dump(ldamodel_obj, os.path.join(base_name, mnb_modelname+".pkl"))
#
# # Load trained model from disk
# ldamodel_obj = joblib.load(os.path.join(base_name, mnb_modelname+".pkl"))
#
# # Print all the topics
# for i,topic in enumerate(ldamodel_obj.print_topics(num_topics=3, num_words=10)):
# #print("topic is {}".format(topic))
# word = topic[1].split("+")
# print("{} : {} ".format(topic[0], word))
#
# # actions are rows and discovered topics are columns
# topic_action_map = np.zeros((real_topic, NUM_TOPICS))
#
# predicted_labels = []
# #vids_list = list(df_train_mag.index)
# for j,vname in enumerate(vids_list):
# label_vid = train_labels[j]
# # sort the tuples with descending topic probabilities
# for index, score in sorted(ldamodel_obj[doc_term_matrix[j]], key=lambda tup: -1*tup[1]):
## for index in [ldamodel_obj[doc_term_matrix[j]].argmax(axis=0)]:
# # print("Score is : {} of Topic: {}".format(score,index))
# #if score>0.5:
# # topic_action_map[label_vid][index]+=1
## score = ldamodel_obj[doc_term_matrix[j]][index]
# topic_action_map[label_vid][index]+=score
# predicted_labels.append(index)
# break
# print("Training Time : topic action mapping is : ")
# print("topic0 topic1 topic2")
# #coloumn are topics and rows are labels
# print(topic_action_map)
# acc_values_tr, perm_tuples_tr, gt_list, pred_list = calculate_accuracy(train_labels,\
# predicted_labels)
# acc_perc = [sum(k)/len(predicted_labels) for k in acc_values_tr]
#
# best_indx = acc_perc.index(max(acc_perc))
# print("Max Acc. : ", max(acc_perc))
# print("Acc values : ", acc_perc)
# print("Acc values : ", acc_values_tr)
# print("perm_tuples : ", perm_tuples_tr)
#model_ang = joblib.load(os.path.join(destpath, mnb_modelname+"_ang.pkl"))
##################################################################################
# Evaluation on validation set
print("Validation phase ....")
if not os.path.isfile(os.path.join(base_name, ft_type + "_feats_val.pkl")):
# features_val, strokes_name_id_val = select_trimmed_feats(c3dFC7FeatsPath, \
# LABELS, val_lst, c3dWinSize)
features_val, strokes_name_id_val = extract_feats(
DATASET,
LABELS,
CLASS_IDS,
BATCH_SIZE,
SEQ_SIZE,
STEP,
extractor=ft_type,
model_path=model_path,
nclasses=NUM_TOPICS,
part='val')
with open(os.path.join(base_name, ft_type + "_feats_val.pkl"),
"wb") as fp:
pickle.dump(features_val, fp)
with open(os.path.join(base_name, ft_type + "_snames_val.pkl"),
"wb") as fp:
pickle.dump(strokes_name_id_val, fp)
else:
with open(os.path.join(base_name, ft_type + "_feats_val.pkl"),
"rb") as fp:
features_val = pickle.load(fp)
with open(os.path.join(base_name, ft_type + "_snames_val.pkl"),
"rb") as fp:
strokes_name_id_val = pickle.load(fp)
print("Create dataframe BOVW validation set...")
df_val_hoof, words_val = create_bovw_df(features_val, strokes_name_id_val, \
km_model, base_name, "val")
vids_list_val = list(df_val_hoof.index)
val_labels = np.array(
[labs_values[labs_keys.index(v)] for v in vids_list_val])
# topic_action_map_val = np.zeros((real_topic, NUM_TOPICS))
# doc_clean_val = [doc.split() for doc in words_val]
# # Creating Dictionary for val set words
# diction_val=corpora.Dictionary(doc_clean_val)
#
# doc_term_matrix_val = [diction_val.doc2bow(doc) for doc in doc_clean_val]
# predicted_label_val = []
# for j,vname in enumerate(vids_list_val):
# label_vid = val_labels[j]
# for index, score in sorted(ldamodel_obj[doc_term_matrix_val[j]], key=lambda tup: -1*tup[1]):
## for index in [ldamodel_obj[doc_term_matrix[j]].argmax(axis=0)]:
## score = ldamodel_obj[doc_term_matrix[j]][index]
# # print("Score is : {} of Topic: {}".format(score,index))
# #if score>0.5:
# # topic_action_map_val[label_vid][index]+=1
# topic_action_map_val[label_vid][index]+=score
# predicted_label_val.append(index)
# break
#
# print(topic_action_map_val)
# labs_df = pd.DataFrame(labels, index=vids_list, columns=['label'])
#
print("Evaluating on the validation set...")
# evaluate(model_mag, df_test_mag, labs_df)
# Find maximum permutation accuracy using predicted_label_val and label_val
# acc_values, perm_tuples, gt_list, pred_list = calculate_accuracy(val_labels, \
# predicted_label_val)
# acc_perc = [sum(k)/len(predicted_label_val) for k in acc_values]
#
# best_indx = acc_perc.index(max(acc_perc))
# print("Max Acc. : ", max(acc_perc))
# print("Acc values : ", acc_perc)
# print("Acc values : ", acc_values)
# print("perm_tuples : ", perm_tuples)
###########################################################################
# Evaluate the BOW classifier (SVM)
confusion_mat = np.zeros((NUM_TOPICS, NUM_TOPICS))
pred = clf.predict(df_val_hoof)
correct = 0
for i, true_val in enumerate(val_labels):
if pred[i] == true_val:
correct += 1
confusion_mat[pred[i], true_val] += 1
print('#' * 30)
print("BOW Classification Results:")
print("%d/%d Correct" % (correct, len(pred)))
print("Accuracy = {} ".format(float(correct) / len(pred)))
print("Confusion matrix")
print(confusion_mat)
return (float(correct) / len(pred))
def main(base_name, traj_len=None, cluster_size=10):
"""
Function to read IDT features, form BOW based model after clustering and evaluate
on 3/5 cluster analysis of highlights dataset.
The videos can be visualized by writing trimmed class videos into their respective
classes.
Parameters:
------
base_name: path to the wts, losses, predictions and log files
"""
seed = 1234
np.random.seed(seed)
print(60 * "#")
#####################################################################
# Divide the sample files into training set, validation and test sets
train_lst, val_lst, test_lst = utils.split_dataset_files(DATASET)
# form the names of the list of label files, should be at destination
train_lab = [f + ".json" for f in train_lst]
val_lab = [f + ".json" for f in val_lst]
test_lab = [f + ".json" for f in test_lst]
# get complete path lists of label files
tr_labs = [os.path.join(LABELS, f) for f in train_lab]
val_labs = [os.path.join(LABELS, f) for f in val_lab]
#####################################################################
sizes = [
utils.getNFrames(os.path.join(DATASET, f + ".avi")) for f in train_lst
]
val_sizes = [
utils.getNFrames(os.path.join(DATASET, f + ".avi")) for f in val_lst
]
###########################################################################
print("No. of training videos : {}".format(len(train_lst)))
print("Size : {}".format(sizes))
# hlDataset = VideoDataset(tr_labs, sizes, seq_size=SEQ_SIZE, is_train_set = True)
#####################################################################
# Feature Extraction : (IDT)
# Get feats for only the training videos. Get ordered histograms of freq
print("Trajectory Length : {}, nClusters : {} ".format(
traj_len, cluster_size))
#####################################################################
# read into dictionary {vidname: np array, ...}
print("Loading features from disk...")
# get Nx4096 numpy matrix with columns as features and rows as window placement features
if not os.path.exists(base_name):
os.makedirs(base_name)
# # Read IDT features {with trajectory length = traj_len}
features, strokes_name_id = read_partition_feats(
DATASET, LABELS, IDT_FEATS + "_TrajLen" + str(traj_len), train_lst,
traj_len)
with open(
os.path.join(base_name,
"idt_feats_traj" + str(traj_len) + ".pkl"),
"wb") as fp:
pickle.dump(features, fp)
with open(
os.path.join(base_name,
"idt_snames_traj" + str(traj_len) + ".pkl"),
"wb") as fp:
pickle.dump(strokes_name_id, fp)
with open(
os.path.join(base_name, "idt_feats_traj" + str(traj_len) + ".pkl"),
"rb") as fp:
features = pickle.load(fp)
with open(
os.path.join(base_name,
"idt_snames_traj" + str(traj_len) + ".pkl"),
"rb") as fp:
strokes_name_id = pickle.load(fp)
#####################################################################
# get small sample of the IDT features and form a matrix (N, vec_size)
vecs = []
for key in sorted(list(features.keys())):
vecs.append(features[key])
vecs = np.vstack(vecs)
vecs[np.isnan(vecs)] = 0
vecs[np.isinf(vecs)] = 0
# sample points for clustering
if vecs.shape[0] > MAX_SAMPLES:
vecs = vecs[
np.random.choice(vecs.shape[0], MAX_SAMPLES, replace=False), :]
#fc7 layer output size (4096)
INP_VEC_SIZE = vecs.shape[-1]
print("INP_VEC_SIZE = ", INP_VEC_SIZE)
km_filepath = os.path.join(base_name, km_filename)
# # Uncomment only while training.
if not os.path.isfile(km_filepath + "_C" + str(cluster_size) + ".pkl"):
km_model = make_codebook(vecs, cluster_size) #, model_type='gmm')
## # Save to disk, if training is performed
print("Writing the KMeans models to disk...")
pickle.dump(
km_model,
open(km_filepath + "_C" + str(cluster_size) + ".pkl", "wb"))
else:
# Load from disk, for validation and test sets.
km_model = pickle.load(
open(km_filepath + "_C" + str(cluster_size) + ".pkl", 'rb'))
###########################################################################
# Form the training dataset for supervised classification
# Assign the words (flow frames) to their closest cluster centres and count the
# frequency for each document(video). Create IDF bow dataframe by weighting
# df_train is (nVids, 50) for magnitude, with index as videonames
print("Create a dataframe for HOOF features...")
df_train, words_train = create_bovw_df(features, strokes_name_id, km_model,\
base_name, "train")
# read the stroke annotation labels from text file.
vids_list = list(df_train.index)
labs_keys, labs_values = get_cluster_labels(ANNOTATION_FILE)
if min(labs_values) == 1:
labs_values = [l - 1 for l in labs_values]
labs_keys = [k.replace('.avi', '') for k in labs_keys]
train_labels = np.array(
[labs_values[labs_keys.index(v)] for v in vids_list])
###########################################################################
# apply_clustering(df_train, DATASET, LABELS, ANNOTATION_FILE, base_name)
###########################################################################
###########################################################################
# Train SVM
clf = LinearSVC(verbose=False, random_state=124, max_iter=3000)
clf.fit(df_train, train_labels)
print("Training Complete.")
###########################################################################
# # Train a classifier on the features.
##################################################################################
# Evaluation on validation set
print("Validation phase ....")
if not os.path.isfile(
os.path.join(base_name,
"idt_feats_val_traj" + str(traj_len) + ".pkl")):
features_val, strokes_name_id_val = read_partition_feats(DATASET, LABELS, \
IDT_FEATS+"_TrajLen"+str(traj_len),
val_lst, traj_len)
with open(
os.path.join(base_name,
"idt_feats_val_traj" + str(traj_len) + ".pkl"),
"wb") as fp:
pickle.dump(features_val, fp)
with open(
os.path.join(base_name,
"idt_snames_val_traj" + str(traj_len) + ".pkl"),
"wb") as fp:
pickle.dump(strokes_name_id_val, fp)
else:
with open(
os.path.join(base_name,
"idt_feats_val_traj" + str(traj_len) + ".pkl"),
"rb") as fp:
features_val = pickle.load(fp)
with open(
os.path.join(base_name,
"idt_snames_val_traj" + str(traj_len) + ".pkl"),
"rb") as fp:
strokes_name_id_val = pickle.load(fp)
print("Create dataframe BOVW validation set...")
df_val_hoof, words_val = create_bovw_df(features_val, strokes_name_id_val, \
km_model, base_name, "val")
vids_list_val = list(df_val_hoof.index)
val_labels = np.array(
[labs_values[labs_keys.index(v)] for v in vids_list_val])
# labs_df = pd.DataFrame(labels, index=vids_list, columns=['label'])
print("Evaluating on the validation set...")
# evaluate(model_mag, df_test_mag, labs_df)
# Find maximum permutation accuracy using predicted_label_val and label_val
# acc_values, perm_tuples, gt_list, pred_list = calculate_accuracy(val_labels, \
# predicted_label_val)
# acc_perc = [sum(k)/len(predicted_label_val) for k in acc_values]
#
# best_indx = acc_perc.index(max(acc_perc))
# print("Max Acc. : ", max(acc_perc))
# print("Acc values : ", acc_perc)
# print("Acc values : ", acc_values)
# print("perm_tuples : ", perm_tuples)
###########################################################################
# Evaluate the BOW classifier (SVM)
confusion_mat = np.zeros((NUM_TOPICS, NUM_TOPICS))
pred = clf.predict(df_val_hoof)
correct = 0
for i, true_val in enumerate(val_labels):
if pred[i] == true_val:
correct += 1
confusion_mat[pred[i], true_val] += 1
print('#' * 30)
print("BOW Classification Results:")
print("%d/%d Correct" % (correct, len(pred)))
print("Accuracy = {} ".format(float(correct) / len(pred)))
print("Confusion matrix")
print(confusion_mat)
return (float(correct) / len(pred))
def main(DATASET,
LABELS,
CLASS_IDS,
BATCH_SIZE,
ANNOTATION_FILE,
SEQ_SIZE=16,
STEP=16,
nstrokes=-1,
N_EPOCHS=25):
'''
Extract sequence features from AutoEncoder.
Parameters:
-----------
DATASET : str
path to the video dataset
LABELS : str
path containing stroke labels
CLASS_IDS : str
path to txt file defining classes, similar to THUMOS
BATCH_SIZE : int
size for batch of clips
SEQ_SIZE : int
no. of frames in a clip (min. 16 for 3D CNN extraction)
STEP : int
stride for next example. If SEQ_SIZE=16, STEP=8, use frames (0, 15), (8, 23) ...
partition : str
'all' / 'train' / 'test' / 'val' : Videos to be considered
nstrokes : int
partial extraction of features (do not execute for entire dataset)
Returns:
--------
trajectories, stroke_names
'''
###########################################################################
attn_utils.seed_everything(1234)
if not os.path.isdir(log_path):
os.makedirs(log_path)
# Read the strokes
# Divide the highlight dataset files into training, validation and test sets
train_lst, val_lst, test_lst = autoenc_utils.split_dataset_files(DATASET)
print("No. of training videos : {}".format(len(train_lst)))
features, stroke_names_id = attn_utils.read_feats(feat_path, feat, snames)
features2, stroke_names_id2 = attn_utils.read_feats(
feat_path2, feat2, snames2)
# get matrix of features from dictionary (N, vec_size)
vecs, vecs2 = [], []
for key in stroke_names_id:
vecs.append(features[key])
vecs2.append(features2[key])
vecs, vecs2 = np.vstack(vecs), np.vstack(vecs2)
vecs[np.isnan(vecs)] = 0
vecs[np.isinf(vecs)] = 0
vecs2[np.isnan(vecs2)] = 0
vecs2[np.isinf(vecs2)] = 0
# vecs = traj_utils.apply_PCA(vecs, 10)
# vecs2 = traj_utils.apply_PCA(vecs2, 10)
# form_lower_dim_dict(features, stroke_names_id, vecs)
# form_lower_dim_dict(features2, stroke_names_id2, vecs2)
#fc7 layer output size (4096)
INP_VEC_SIZE, INP_VEC_SIZE2 = vecs.shape[-1], vecs2.shape[-1]
print("INP_VEC_SIZE = {} : INP_VEC_SIZE2 = {}".format(
INP_VEC_SIZE, INP_VEC_SIZE2))
km_filepath = os.path.join(log_path, km_filename)
# Feats1
if not os.path.isfile(km_filepath + "_C" + str(cluster_size) + ".pkl"):
km_model = make_codebook(vecs, cluster_size) #, model_type='gmm')
## # Save to disk, if training is performed
print("Writing the KMeans models to disk...")
pickle.dump(
km_model,
open(km_filepath + "_C" + str(cluster_size) + ".pkl", "wb"))
else:
# Load from disk, for validation and test sets.
km_model = pickle.load(
open(km_filepath + "_C" + str(cluster_size) + ".pkl", 'rb'))
# Feats2
if not os.path.isfile(km_filepath + "_C" + str(cluster_size) + "_2.pkl"):
km_model2 = make_codebook(vecs2, cluster_size) #, model_type='gmm')
## # Save to disk, if training is performed
print("Writing the KMeans models to disk...")
pickle.dump(
km_model2,
open(km_filepath + "_C" + str(cluster_size) + "_2.pkl", "wb"))
else:
# Load from disk, for validation and test sets.
km_model2 = pickle.load(
open(km_filepath + "_C" + str(cluster_size) + "_2.pkl", 'rb'))
print("Create numpy one hot representation for train features...")
onehot_feats = create_bovw_SA(features, stroke_names_id, km_model)
print("Create numpy one hot representation for train features2...")
onehot_feats2 = create_bovw_SA(features2, stroke_names_id2, km_model2)
ft_path = os.path.join(log_path,
"onehot_C" + str(cluster_size) + "_train.pkl")
ft_path2 = os.path.join(log_path,
"onehot_C" + str(cluster_size) + "_train_2.pkl")
with open(ft_path, "wb") as fp:
pickle.dump(onehot_feats, fp)
with open(ft_path2, "wb") as fp:
pickle.dump(onehot_feats2, fp)
###########################################################################
features_val, stroke_names_id_val = attn_utils.read_feats(
feat_path, feat_val, snames_val)
features_val2, stroke_names_id_val2 = attn_utils.read_feats(
feat_path2, feat_val2, snames_val2)
# # get matrix of features from dictionary (N, vec_size)
# vecs, vecs2 = [], []
# for key in stroke_names_id:
# vecs.append(features[key])
# vecs2.append(features2[key])
# vecs, vecs2 = np.vstack(vecs), np.vstack(vecs2)
#
# vecs[np.isnan(vecs)] = 0
# vecs[np.isinf(vecs)] = 0
# vecs2[np.isnan(vecs2)] = 0
# vecs2[np.isinf(vecs2)] = 0
#
# form_lower_dim_dict(features, stroke_names_id, vecs)
# form_lower_dim_dict(features2, stroke_names_id2, vecs2)
print("Create numpy one hot representation for val features...")
onehot_feats_val = create_bovw_SA(features_val, stroke_names_id_val,
km_model)
print("Create numpy one hot representation for val features2...")
onehot_feats_val2 = create_bovw_SA(features_val2, stroke_names_id_val2,
km_model2)
ft_path_val = os.path.join(log_path,
"onehot_C" + str(cluster_size) + "_val.pkl")
ft_path_val2 = os.path.join(log_path,
"onehot_C" + str(cluster_size) + "_val_2.pkl")
with open(ft_path_val, "wb") as fp:
pickle.dump(onehot_feats_val, fp)
with open(ft_path_val2, "wb") as fp:
pickle.dump(onehot_feats_val2, fp)
###########################################################################
# Create a Dataset
# ft_path = os.path.join(base_name, ft_dir, feat)
train_dataset = StrokeFeatureSequencesDataset(ft_path,
ft_path2,
train_lst,
DATASET,
LABELS,
CLASS_IDS,
frames_per_clip=SEQ_SIZE,
extracted_frames_per_clip=2,
step_between_clips=STEP,
train=True)
# ft_path_val = os.path.join(base_name, ft_dir, feat_val)
val_dataset = StrokeFeatureSequencesDataset(ft_path_val,
ft_path_val2,
val_lst,
DATASET,
LABELS,
CLASS_IDS,
frames_per_clip=SEQ_SIZE,
extracted_frames_per_clip=2,
step_between_clips=STEP,
train=False)
# get labels
labs_keys, labs_values = attn_utils.get_cluster_labels(ANNOTATION_FILE)
# created weighted Sampler for class imbalance
samples_weight = attn_utils.get_sample_weights(train_dataset, labs_keys,
labs_values, train_lst)
sampler = WeightedRandomSampler(samples_weight, len(samples_weight))
train_loader = DataLoader(dataset=train_dataset,
batch_size=BATCH_SIZE,
sampler=sampler,
worker_init_fn=np.random.seed(12))
val_loader = DataLoader(dataset=val_dataset,
batch_size=BATCH_SIZE,
shuffle=False)
data_loaders = {"train": train_loader, "test": val_loader}
num_classes = len(list(set(labs_values)))
# vis_clusters(features, onehot_feats, stroke_names_id, 2, DATASET, log_path)
###########################################################################
# load model and set loss function
model = attn_model.GRUBoWMultiStreamClassifier(INPUT_SIZE, INPUT_SIZE,
HIDDEN_SIZE, HIDDEN_SIZE,
num_classes, N_LAYERS,
bidirectional)
# model = load_weights(base_name, model, N_EPOCHS, "Adam")
# for ft in model.parameters():
# ft.requires_grad = False
# Setup the loss fxn
criterion = nn.CrossEntropyLoss()
model = model.to(device)
print("Params to learn:")
params_to_update = []
for name, param in model.named_parameters():
if param.requires_grad == True:
params_to_update.append(param)
# print("\t",name)
# Observe that all parameters are being optimized
optimizer_ft = torch.optim.Adam(model.parameters(), lr=0.001)
# optimizer_ft = torch.optim.SGD(model.parameters(), lr=0.01, momentum=0.9)
# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = StepLR(optimizer_ft, step_size=15, gamma=0.1)
###########################################################################
# Training the model
start = time.time()
model = train_model(features,
stroke_names_id,
model,
data_loaders,
criterion,
optimizer_ft,
exp_lr_scheduler,
labs_keys,
labs_values,
num_epochs=N_EPOCHS)
end = time.time()
#
# # save the best performing model
attn_utils.save_model_checkpoint(
log_path, model, N_EPOCHS,
"S" + str(SEQ_SIZE) + "C" + str(cluster_size) + "_SGD")
# Load model checkpoints
model = attn_utils.load_weights(
log_path, model, N_EPOCHS,
"S" + str(SEQ_SIZE) + "C" + str(cluster_size) + "_SGD")
print("Total Execution time for {} epoch : {}".format(
N_EPOCHS, (end - start)))
# ###########################################################################
acc = predict(features_val,
stroke_names_id_val,
model,
data_loaders,
labs_keys,
labs_values,
SEQ_SIZE,
phase='test')
# call count_paramters(model) for displaying total no. of parameters
print("#Parameters : {} ".format(autoenc_utils.count_parameters(model)))
return acc
def main(DATASET, LABELS, SEQ_SIZE=16, STEP=16, nstrokes=-1):
'''
Extract sequence features from AutoEncoder.
Parameters:
-----------
DATASET : str
path to the video dataset
LABELS : str
path containing stroke labels
BATCH_SIZE : int
size for batch of clips
SEQ_SIZE : int
no. of frames in a clip (min. 16 for 3D CNN extraction)
STEP : int
stride for next example. If SEQ_SIZE=16, STEP=8, use frames (0, 15), (8, 23) ...
partition : str
'all' / 'train' / 'test' / 'val' : Videos to be considered
nstrokes : int
partial extraction of features (do not execute for entire dataset)
Returns:
--------
trajectories, stroke_names
'''
###########################################################################
# attn_utils.seed_everything(1234)
if not os.path.isdir(log_path):
os.makedirs(log_path)
hmdb51_train = hmdb.HMDB51(DATASET,
LABELS,
SEQ_SIZE,
step_between_clips=STEP,
fold=1,
train=True,
transform=None)
hmdb51_test = hmdb.HMDB51(DATASET,
LABELS,
SEQ_SIZE,
step_between_clips=STEP,
fold=1,
train=False,
transform=None)
# # Extracting training features
# extract_of_features(feat_path, hmdb51_train.video_list, hmdb51_train.fold, hmdb51_train.train)
# # Extracting testing features
# extract_of_features(feat_path, hmdb51_test.video_list, hmdb51_test.fold, hmdb51_test.train)
features, stroke_names_id = read_feats(feat_path, feat, snames)
# get matrix of features from dictionary (N, vec_size)
vecs = []
for key in sorted(list(features.keys())):
vecs.append(features[key])
vecs = np.vstack(vecs)
vecs[np.isnan(vecs)] = 0
vecs[np.isinf(vecs)] = 0
#fc7 layer output size (4096)
INP_VEC_SIZE = vecs.shape[-1]
print("INP_VEC_SIZE = ", INP_VEC_SIZE)
km_filepath = os.path.join(log_path, km_filename)
# # Uncomment only while training.
if not os.path.isfile(km_filepath + "_C" + str(cluster_size) + ".pkl"):
km_model = make_codebook(vecs, cluster_size) #, model_type='gmm')
## # Save to disk, if training is performed
print("Writing the KMeans models to disk...")
pickle.dump(
km_model,
open(km_filepath + "_C" + str(cluster_size) + ".pkl", "wb"))
else:
# Load from disk, for validation and test sets.
km_model = pickle.load(
open(km_filepath + "_C" + str(cluster_size) + ".pkl", 'rb'))
df_train, words_train = create_bovw_df(features, stroke_names_id, km_model,
log_path, 'train')
# read the stroke annotation labels from text file.
vids_list = list(df_train.index)
labs_keys = hmdb51_train.video_list
labs_indx = hmdb51_train.indices
labs_values = [hmdb51_train.samples[i][1] for i in labs_indx]
train_labels = np.array(
[labs_values[labs_keys.index(v)] for v in vids_list])
num_classes = len(list(set(labs_values)))
print("Training dataframe formed.")
###########################################################################
# Train a classifier on the features.
###########################################################################
# Train SVM
clf = LinearSVC(verbose=False, random_state=124, max_iter=3000)
clf.fit(df_train, train_labels)
print("Training Complete.")
###########################################################################
#print("Training complete. Saving to disk.")
# Save model to disk
joblib.dump(clf, os.path.join(log_path, "clf.pkl"))
# Load trained model from disk
clf = joblib.load(os.path.join(log_path, "clf.pkl"))
# Train a classifier on both the features.
#print("Training with SVM")
#df_train = pd.concat([df_train_mag, df_train_ang], axis=1)
#clf_both = SVC(kernel="linear",verbose=True)
#clf_both = LinearSVC(verbose=True, random_state=123, max_iter=2000)
#clf_both.fit(df_train, labels)
#print("Training with SVM (ang)")
#clf_ang = SVC(kernel="linear",verbose=True)
#clf_ang.fit(df_train_ang, labels)
##########################################################################
features_test, stroke_names_id_test = read_feats(feat_path, feat_test,
snames_test)
print("Create dataframe BOVW validation set...")
df_test_hoof, words_test = create_bovw_df(features_test, stroke_names_id_test, \
km_model, log_path, "test")
vids_list_test = list(df_test_hoof.index)
labs_keys = hmdb51_test.video_list
labs_indx = hmdb51_test.indices
labs_values = [hmdb51_test.samples[i][1] for i in labs_indx]
test_labels = np.array(
[labs_values[labs_keys.index(v)] for v in vids_list_test])
###########################################################################
# Evaluate the BOW classifier (SVM)
confusion_mat = np.zeros((num_classes, num_classes))
pred = clf.predict(df_test_hoof)
correct = 0
for i, true_val in enumerate(test_labels):
if pred[i] == true_val:
correct += 1
confusion_mat[pred[i], true_val] += 1
print('#' * 30)
print("BOW Classification Results:")
print("%d/%d Correct" % (correct, len(pred)))
print("Accuracy = {} ".format(float(correct) / len(pred)))
print("Confusion matrix")
print(confusion_mat)
return (float(correct) / len(pred))
with open(os.path.join(base_name, "feats.pkl"), "rb") as fp:
features = pickle.load(fp)
#
#####################################################################
# get matrix of features from dictionary (N, vec_size)
vecs = []
for key in sorted(list(features.keys())):
vecs.append(features[key])
vecs = np.vstack(vecs)
#fc7 layer output size (4096)
INP_VEC_SIZE = vecs.shape[-1]
print("INP_VEC_SIZE = ", INP_VEC_SIZE)
km_filepath = os.path.join(base_name, km_filename+".pkl")
# # Uncomment only while training.
km_model = make_codebook(vecs, cluster_size)
# # Save to disk, if training is performed
# print("Writing the KMeans models to disk...")
pickle.dump(km_model, open(km_filepath, "wb"))
# Load from disk, for validation and test sets.
km_model = pickle.load(open(km_filepath, 'rb'))
###########################################################################
# Form the training dataset for supervised classification
# Assign the words (flow frames) to their closest cluster centres and count the
# frequency for each document(video). Create IDF bow dataframe by weighting
# df_train is (nVids, 50) for magnitude, with index as videonames
# print("Create a dataframe for C3D FC7 features...")
# df_train_c3d, words_train = create_bovw_c3d_traindf(features, \
# strokes_name_id, km_model, c3dWinSize)
def main(DATASET,
LABELS,
BATCH_SIZE,
SEQ_SIZE=16,
STEP=16,
nstrokes=-1,
N_EPOCHS=25):
'''
Extract sequence features from AutoEncoder.
Parameters:
-----------
DATASET : str
path to the video dataset
LABELS : str
path containing stroke labels
BATCH_SIZE : int
size for batch of clips
SEQ_SIZE : int
no. of frames in a clip (min. 16 for 3D CNN extraction)
STEP : int
stride for next example. If SEQ_SIZE=16, STEP=8, use frames (0, 15), (8, 23) ...
partition : str
'all' / 'train' / 'test' / 'val' : Videos to be considered
nstrokes : int
partial extraction of features (do not execute for entire dataset)
Returns:
--------
trajectories, stroke_names
'''
###########################################################################
attn_utils.seed_everything(1234)
if not os.path.isdir(log_path):
os.makedirs(log_path)
features, stroke_names_id = attn_utils.read_feats(feat_path, feat, snames)
# get matrix of features from dictionary (N, vec_size)
vecs = []
for key in sorted(list(features.keys())):
vecs.append(features[key])
vecs = np.vstack(vecs)
vecs[np.isnan(vecs)] = 0
vecs[np.isinf(vecs)] = 0
#fc7 layer output size (4096)
INP_VEC_SIZE = vecs.shape[-1]
print("INP_VEC_SIZE = ", INP_VEC_SIZE)
km_filepath = os.path.join(log_path, km_filename)
# # Uncomment only while training.
if not os.path.isfile(km_filepath + "_C" + str(cluster_size) + ".pkl"):
km_model = make_codebook(vecs, cluster_size) #, model_type='gmm')
## # Save to disk, if training is performed
print("Writing the KMeans models to disk...")
pickle.dump(
km_model,
open(km_filepath + "_C" + str(cluster_size) + ".pkl", "wb"))
else:
# Load from disk, for validation and test sets.
km_model = pickle.load(
open(km_filepath + "_C" + str(cluster_size) + ".pkl", 'rb'))
print("Create numpy one hot representation for train features...")
onehot_feats = create_bovw_onehot(features, stroke_names_id, km_model)
ft_path = os.path.join(log_path, "C" + str(cluster_size) + "_train.pkl")
# ft_path = os.path.join(feat_path, feat)
with open(ft_path, "wb") as fp:
pickle.dump(onehot_feats, fp)
with open(
os.path.join(log_path,
"C" + str(cluster_size) + "_snames_train.pkl"),
"wb") as fp:
pickle.dump(stroke_names_id, fp)
#########################################################################
#########################################################################
features_test, stroke_names_id_test = attn_utils.read_feats(
feat_path, feat_test, snames_test)
print("Create numpy one hot representation for val features...")
onehot_feats_test = create_bovw_onehot(features_test, stroke_names_id_test,
km_model)
ft_path_test = os.path.join(log_path,
"C" + str(cluster_size) + "_test.pkl")
# ft_path_test = os.path.join(feat_path, feat_test)
with open(ft_path_test, "wb") as fp:
pickle.dump(onehot_feats_test, fp)
with open(
os.path.join(log_path,
"C" + str(cluster_size) + "_snames_test.pkl"),
"wb") as fp:
pickle.dump(stroke_names_id_test, fp)
###########################################################################
# Create a Dataset
train_dataset = hmdb.HMDB51FeatureSequenceDataset(
ft_path,
DATASET,
LABELS,
frames_per_clip=SEQ_SIZE,
extracted_frames_per_clip=16,
step_between_clips=STEP,
train=True)
test_dataset = hmdb.HMDB51FeatureSequenceDataset(
ft_path_test,
DATASET,
LABELS,
frames_per_clip=SEQ_SIZE,
extracted_frames_per_clip=16,
step_between_clips=STEP,
train=False)
# display_sizes(train_dataset.video_list)
# display_sizes(test_dataset.video_list)
# # created weighted Sampler for class imbalance
samples_weight = get_hmdb_sample_weights(train_dataset)
sampler = WeightedRandomSampler(samples_weight, len(samples_weight))
train_loader = DataLoader(
dataset=train_dataset,
batch_size=BATCH_SIZE, #shuffle=True)
sampler=sampler,
worker_init_fn=np.random.seed(12))
test_loader = DataLoader(dataset=test_dataset,
batch_size=BATCH_SIZE,
shuffle=False)
data_loaders = {"train": train_loader, "test": test_loader}
# num_classes = len(list(set(labs_values)))
num_classes = 51
###########################################################################
# load model and set loss function
model = attn_model.GRUBoWHAClassifier(INPUT_SIZE, HIDDEN_SIZE, num_classes,
N_LAYERS, bidirectional)
# model = load_weights(base_name, model, N_EPOCHS, "Adam")
# for ft in model.parameters():
# ft.requires_grad = False
# Setup the loss fxn
criterion = nn.CrossEntropyLoss()
model = model.to(device)
# print("Params to learn:")
params_to_update = []
for name, param in model.named_parameters():
if param.requires_grad == True:
params_to_update.append(param)
# print("\t",name)
# Observe that all parameters are being optimized
optimizer_ft = torch.optim.Adam(model.parameters(), lr=0.001)
# optimizer_ft = torch.optim.SGD(model.parameters(), lr=0.01, momentum=0.9)
# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = StepLR(optimizer_ft, step_size=10, gamma=0.1)
###########################################################################
# Training the model
start = time.time()
model = train_model(model,
data_loaders,
criterion,
optimizer_ft,
exp_lr_scheduler,
num_epochs=N_EPOCHS)
end = time.time()
# save the best performing model
attn_utils.save_model_checkpoint(
log_path, model, N_EPOCHS,
"S" + str(SEQ_SIZE) + "C" + str(cluster_size) + "_SGD")
# Load model checkpoints
model = attn_utils.load_weights(
log_path, model, N_EPOCHS,
"S" + str(SEQ_SIZE) + "C" + str(cluster_size) + "_SGD")
print("Total Execution time for {} epoch : {}".format(
N_EPOCHS, (end - start)))
###########################################################################
acc = predict(model, data_loaders, SEQ_SIZE, phase='test')
# call count_paramters(model) for displaying total no. of parameters
print("#Parameters : {} ".format(autoenc_utils.count_parameters(model)))
return acc
