def load_models(opt):
opt.resume_path = opt.resume_path_det
opt.pretrain_path = opt.pretrain_path_det
opt.sample_duration = opt.sample_duration_det
opt.model = opt.model_det
opt.model_depth = opt.model_depth_det
opt.modality = opt.modality_det
opt.resnet_shortcut = opt.resnet_shortcut_det
opt.n_classes = opt.n_classes_det
opt.n_finetune_classes = opt.n_finetune_classes_det
if opt.root_path != '':
opt.video_path = os.path.join(opt.root_path, opt.video_path)
opt.annotation_path = os.path.join(opt.root_path, opt.annotation_path)
opt.result_path = os.path.join(opt.root_path, opt.result_path)
if opt.resume_path:
opt.resume_path = os.path.join(opt.root_path, opt.resume_path)
if opt.pretrain_path:
opt.pretrain_path = os.path.join(opt.root_path, opt.pretrain_path)
opt.scales = [opt.initial_scale]
for i in range(1, opt.n_scales):
opt.scales.append(opt.scales[-1] * opt.scale_step)
opt.arch = '{}-{}'.format(opt.model, opt.model_depth)
opt.mean = get_mean(opt.norm_value)
opt.std = get_std(opt.norm_value)
print(opt)
with open(os.path.join(opt.result_path, 'opts_det.json'), 'w') as opt_file:
json.dump(vars(opt), opt_file)
torch.manual_seed(opt.manual_seed)
detector, parameters = generate_model(opt)
if opt.resume_path:
print('loading checkpoint {}'.format(opt.resume_path))
checkpoint = torch.load(opt.resume_path)
assert opt.arch == checkpoint['arch']
detector.load_state_dict(checkpoint['state_dict'])
print('Model 1 \n', detector)
pytorch_total_params = sum(p.numel() for p in detector.parameters()
if p.requires_grad)
print("Total number of trainable parameters: ", pytorch_total_params)
# Reset parsed args
opt = parse_opts_online()
opt.resume_path = opt.resume_path_clf
opt.pretrain_path = opt.pretrain_path_clf
opt.sample_duration = opt.sample_duration_clf
opt.model = opt.model_clf
opt.model_depth = opt.model_depth_clf
opt.modality = opt.modality_clf
opt.resnet_shortcut = opt.resnet_shortcut_clf
opt.n_classes = opt.n_classes_clf
opt.n_finetune_classes = opt.n_finetune_classes_clf
if opt.root_path != '':
opt.video_path = os.path.join(opt.root_path, opt.video_path)
opt.annotation_path = os.path.join(opt.root_path, opt.annotation_path)
opt.result_path = os.path.join(opt.root_path, opt.result_path)
if opt.resume_path:
opt.resume_path = os.path.join(opt.root_path, opt.resume_path)
if opt.pretrain_path:
opt.pretrain_path = os.path.join(opt.root_path, opt.pretrain_path)
opt.scales = [opt.initial_scale]
for i in range(1, opt.n_scales):
opt.scales.append(opt.scales[-1] * opt.scale_step)
if opt.model == 'ssar':
opt.arch = opt.model
else:
opt.arch = '{}-{}'.format(opt.model, opt.model_depth)
opt.mean = get_mean(opt.norm_value)
opt.std = get_std(opt.norm_value)
print(opt)
with open(os.path.join(opt.result_path, 'opts_clf.json'), 'w') as opt_file:
json.dump(vars(opt), opt_file)
torch.manual_seed(opt.manual_seed)
classifier, parameters = generate_model(opt)
if opt.resume_path:
print('loading pretrained model {}'.format(opt.pretrain_path))
checkpoint = torch.load(opt.resume_path)
assert opt.arch == checkpoint['arch']
classifier.load_state_dict(checkpoint['state_dict'])
print('Model 2 \n', classifier)
pytorch_total_params = sum(p.numel() for p in classifier.parameters()
if p.requires_grad)
print("Total number of trainable parameters: ", pytorch_total_params)
return detector, classifier
from temporal_transforms import *
from target_transforms import ClassLabel
from dataset import get_online_data
from utils import AverageMeter, LevenshteinDistance, Queue
import pdb
import time
import numpy as np
import datetime
def weighting_func(x):
return (1 / (1 + np.exp(-0.2 * (x - 9))))
opt = parse_opts_online()
opt = parse_opts_online()
def load_models(opt):
opt.resume_path = opt.resume_path_clf
opt.pretrain_path = opt.pretrain_path_clf
opt.sample_duration = opt.sample_duration_clf
opt.model = opt.model_clf
opt.model_depth = opt.model_depth_clf
opt.width_mult = opt.width_mult_clf
opt.modality = opt.modality_clf
opt.resnet_shortcut = opt.resnet_shortcut_clf
opt.n_classes = opt.n_classes_clf
opt.n_finetune_classes = opt.n_finetune_classes_clf
def main(clf_threshold_pre):
print(f'Early-detection threshold: {clf_threshold_pre}')
opt = parse_opts_online()
detector, classifier = load_models(opt)
if opt.no_mean_norm and not opt.std_norm:
norm_method = Normalize([0, 0, 0], [1, 1, 1])
elif not opt.std_norm:
norm_method = Normalize(opt.mean, [1, 1, 1])
else:
norm_method = Normalize(opt.mean, opt.std)
if opt.model_clf == 'ssar':
opt.sample_size_clf = (126, 224)
opt.mean_clf = (0.485, 0.456, 0.406)
opt.std_clf = (0.229, 0.224, 0.225)
spatial_transform_clf = transforms.Compose([
transforms.Resize(opt.sample_size_clf),
transforms.ToTensor(),
transforms.Normalize(opt.mean_clf, opt.std_clf)
])
spatial_transform = Compose(
[Scale(112),
CenterCrop(112),
ToTensor(opt.norm_value), norm_method])
target_transform = ClassLabel()
## Get list of videos to test
if opt.dataset == 'egogesture':
subject_list = [
'Subject{:02d}'.format(i)
for i in [2, 9, 11, 14, 18, 19, 28, 31, 41, 47]
]
test_paths = []
for subject in subject_list:
for x in glob.glob(
os.path.join(opt.video_path, subject, '*/*/rgb*/')):
test_paths.append(x)
elif opt.dataset == 'nv':
df = pd.read_csv(os.path.join(
opt.video_path, 'nvgesture_test_correct_cvpr2016_v2.lst'),
delimiter=' ',
header=None)
test_paths = []
for x in df[0].values:
test_paths.append(
os.path.join(opt.video_path, x.replace('path:', ''),
'sk_color_all').replace(os.sep, '/'))
# Figures setup
# fig, ax = plt.subplots(nrows=6, ncols=1)
# x_data, y_datas = [], []
# lines = []
# for j in range(6):
# if j != 0:
# ax[j].set_xlim(0, 400)
# ax[j].set_ylim(0, 1)
# y_datas.append([])
# lines.append([])
# for _ in range(opt.n_classes_clf):
# y_data = []
# y_datas[j].append(y_data)
# line, = ax[j].plot(x_data, y_data)
# lines[j].append(line)
print('Start Evaluation')
detector.eval()
classifier.eval()
levenshtein_accuracies = AverageMeter()
frames_early_meter = AverageMeter()
videoidx = 0
for path in test_paths[4:]:
path = os.path.normpath(path)
if opt.dataset == 'egogesture':
opt.whole_path = path.rsplit(os.sep, 4)[-4:]
opt.whole_path = os.sep.join(opt.whole_path)
elif opt.dataset == 'nv':
opt.whole_path = path.split(
os.sep,
3) # TODO: fix bad dependency on fixed depth file locations
opt.whole_path = opt.whole_path[-1]
videoidx += 1
active_index = 0
passive_count = 999
active = False
prev_active = False
finished_prediction = None
pre_predict = False
cum_sum = np.zeros(opt.n_classes_clf, )
cum_sum_unweighted = np.zeros(opt.n_classes_clf, )
clf_selected_queue = np.zeros(opt.n_classes_clf, )
det_selected_queue = np.zeros(opt.n_classes_det, )
myqueue_det = Queue(opt.det_queue_size, n_classes=opt.n_classes_det)
myqueue_clf = Queue(opt.clf_queue_size, n_classes=opt.n_classes_clf)
print('[{}/{}]============'.format(videoidx, len(test_paths)))
print(path)
opt.sample_duration = max(opt.sample_duration_clf,
opt.sample_duration_det)
if opt.model_clf == 'ssar':
test_data = get_online_data(
opt, [spatial_transform, spatial_transform_clf],
None,
target_transform,
modality='RGB')
else:
test_data = get_online_data(opt, spatial_transform, None,
target_transform)
test_loader = torch.utils.data.DataLoader(test_data,
batch_size=opt.batch_size,
shuffle=False,
num_workers=opt.n_threads,
pin_memory=True)
results = []
prev_best1 = opt.n_classes_clf
if opt.model_clf == 'ssar':
# Init recurrent state zero
lstm_hidden = [None, None, None, None]
for i, (inputs, targets) in enumerate(test_loader):
if opt.model_clf == 'ssar':
inputs, inputs_clf = inputs
if not opt.no_cuda:
targets = targets.cuda(non_blocking=True)
ground_truth_array = np.zeros(opt.n_classes_clf + 1, )
with torch.no_grad():
inputs = Variable(inputs)
targets = Variable(targets)
if opt.modality_det == 'RGB':
inputs_det = inputs[:, :3, -opt.sample_duration_det:, :, :]
elif opt.modality_det == 'Depth':
inputs_det = inputs[:, -1, -opt.
sample_duration_det:, :, :].unsqueeze(
1)
elif opt.modality_det == 'RGB-D':
inputs_det = inputs[:, :, -opt.sample_duration_det:, :, :]
# print(inputs_det[0, :, -1, 0:4, 0:4])
outputs_det = detector(inputs_det)
outputs_det = F.softmax(outputs_det, dim=1)
outputs_det = outputs_det.cpu().numpy()[0].reshape(-1, )
# enqueue the probabilities to the detector queue
myqueue_det.enqueue(outputs_det.tolist())
if opt.det_strategy == 'raw':
det_selected_queue = outputs_det
elif opt.det_strategy == 'median':
det_selected_queue = myqueue_det.median
elif opt.det_strategy == 'ma':
det_selected_queue = myqueue_det.ma
elif opt.det_strategy == 'ewma':
det_selected_queue = myqueue_det.ewma
prediction_det = np.argmax(det_selected_queue)
prob_det = det_selected_queue[1]
#### State of the detector is checked here as detector act as a switch for the classifier
if prediction_det == 1:
if opt.model_clf == 'ssar':
inputs_clf = Variable(inputs_clf)
if not opt.no_cuda:
inputs_clf = inputs_clf.cuda()
if opt.modality_clf == 'RGB':
inputs_clf = inputs_clf[:, :3, -1, :, :]
elif opt.modality_clf == 'Depth':
inputs_clf = inputs_clf[:, -1,
-1, :, :].unsqueeze(1)
elif opt.modality_clf == 'RGB-D':
inputs_clf = inputs_clf[:, :, -1, :, :]
outputs_clf, lstm_hidden = classifier(
inputs_clf, lstm_hidden, get_lstm_state=True)
else:
if opt.modality_clf == 'RGB':
inputs_clf = inputs[:, :3, :, :, :]
elif opt.modality_clf == 'Depth':
inputs_clf = inputs[:, -1, :, :, :].unsqueeze(1)
elif opt.modality_clf == 'RGB-D':
inputs_clf = inputs[:, :, :, :, :]
outputs_clf = classifier(inputs_clf)
outputs_clf = F.softmax(outputs_clf, dim=1)
outputs_clf = outputs_clf.cpu().numpy()[0].reshape(-1, )
# Push the probabilities to queue
myqueue_clf.enqueue(outputs_clf.tolist())
passive_count = 0
if opt.clf_strategy == 'raw':
clf_selected_queue = outputs_clf
elif opt.clf_strategy == 'median':
clf_selected_queue = myqueue_clf.median
elif opt.clf_strategy == 'ma':
clf_selected_queue = myqueue_clf.ma
elif opt.clf_strategy == 'ewma':
clf_selected_queue = myqueue_clf.ewma
else:
if opt.model_clf == 'ssar':
# Reset recurrent state
lstm_hidden = [None, None, None, None]
outputs_clf = np.zeros(opt.n_classes_clf, )
# Push the probabilities to queue
myqueue_clf.enqueue(outputs_clf.tolist())
passive_count += 1
if passive_count >= opt.det_counter:
active = False
else:
active = True
# one of the following line need to be commented !!!!
if active:
active_index += 1
cum_sum = ((cum_sum * (active_index - 1)) +
(weighting_func(active_index) * clf_selected_queue)
) / active_index # Weighted Aproach
cum_sum_unweighted = ((cum_sum_unweighted *
(active_index - 1)) +
(1.0 * clf_selected_queue)
) / active_index #Not Weighting Aproach
best2, best1 = tuple(cum_sum.argsort()[-2:][::1])
if float(cum_sum[best1] - cum_sum[best2]) > clf_threshold_pre:
finished_prediction = True
pre_predict = True
else:
active_index = 0
# Visualize
# x_data.append(i)
# y_datas[1][0].append(prob_det)
# lines[1][0].set_xdata(x_data)
# lines[1][0].set_ydata(y_datas[1][0])
# for j in range(opt.n_classes_clf):
# y_datas[2][j].append(cum_sum[j])
# y_datas[3][j].append(cum_sum_unweighted[j])
# y_datas[4][j].append(clf_selected_queue[j] if active else 0)
# for k in range(2, 5):
# lines[k][j].set_xdata(x_data)
# lines[k][j].set_ydata(y_datas[k][j])
# for k in range(1, 6):
# ax[k].set_xlim(i - 400, i)
# mean = np.array(opt.mean, dtype=np.float32).reshape(1, 1, -1)
# img = inputs_det[0, :, -1].permute(1, 2, 0).cpu().numpy() + mean
# img = img.astype(int)
# if i == 0:
# im_plt = ax[0].imshow(img)
# else:
# im_plt.set_data(img)
if i % 10 == 0:
plt.draw()
plt.pause(0.001)
if active == False and prev_active == True:
finished_prediction = True
elif active == True and prev_active == False:
finished_prediction = False
if finished_prediction == True:
detection_frame = (i *
opt.stride_len) + opt.sample_duration_clf
best2, best1 = tuple(cum_sum.argsort()[-2:][::1])
if cum_sum[best1] > opt.clf_threshold_final:
if pre_predict == True:
if best1 != prev_best1:
if cum_sum[best1] > opt.clf_threshold_final:
results.append((detection_frame, best1))
print(
'Early Detected - class : {} with prob : {} at frame {}'
.format(best1, cum_sum[best1],
detection_frame))
else:
# raw_best = clf_selected_queue.argsort()[-1]
# results.append((detection_frame,raw_best))
# print( 'Late Detected - class : {} with prob : {} at frame {}'.format(raw_best, clf_selected_queue[raw_best], detection_frame))
if cum_sum[best1] > opt.clf_threshold_final:
if best1 == prev_best1:
if cum_sum[best1] > 5:
results.append((detection_frame, best1))
print(
'Late Detected - class : {} with prob : {} at frame {}'
.format(best1, cum_sum[best1],
detection_frame))
else:
results.append((detection_frame, best1))
print(
'Late Detected - class : {} with prob : {} at frame {}'
.format(best1, cum_sum[best1],
detection_frame))
prev_best1 = best1
finished_prediction = False
# prev_best1 = best1
# finished_prediction = False
cum_sum = np.zeros(opt.n_classes_clf, )
cum_sum_unweighted = np.zeros(opt.n_classes_clf, )
if active == False and prev_active == True:
pre_predict = False
prev_active = active
if opt.dataset == 'egogesture':
opt.video_path = os.path.normpath(opt.video_path)
opt.whole_path = os.path.normpath(opt.whole_path)
target_csv_path = os.path.join(
opt.video_path, 'labels-final-revised1',
opt.whole_path.rsplit(os.sep, 2)[0],
'Group' + opt.whole_path.rsplit('.', 1)[0][-1] +
'.csv').replace('Subject', 'subject')
true_classes = []
end_frames = []
with open(target_csv_path) as csvfile:
readCSV = csv.reader(csvfile, delimiter=',')
for row in readCSV:
true_classes.append(int(row[0]) - 1)
end_frames.append(int(row[2]))
elif opt.dataset == 'nv':
true_classes = []
with open('./annotation_nvGesture/vallistall.txt') as csvfile:
readCSV = csv.reader(csvfile, delimiter=' ')
for row in readCSV:
if row[0] == opt.whole_path:
if row[1] != '26':
true_classes.append(int(row[1]) - 1)
predicted = np.array(results)[:, 1]
detection_frames = np.array(results)[:, 0]
true_classes = np.array(true_classes)
levenshtein_distance, avg_frames_early = LevenshteinDistancePlusAvgFramesEarly(
true_classes, predicted, end_frames, detection_frames)
levenshtein_accuracy = 1 - (levenshtein_distance / len(true_classes))
if levenshtein_distance < 0: # Distance cannot be less than 0
levenshtein_accuracies.update(0, len(true_classes))
else:
levenshtein_accuracies.update(levenshtein_accuracy,
len(true_classes))
frames_early_meter.update(avg_frames_early)
print('predicted classes: \t', predicted)
print('True classes :\t\t', true_classes)
print('Levenshtein Accuracy = {} ({})'.format(
levenshtein_accuracies.val, levenshtein_accuracies.avg))
print(
f'Average frames early = {frames_early_meter.val} ({frames_early_meter.avg})'
)
print('Average Levenshtein Accuracy= {}'.format(
levenshtein_accuracies.avg))
print('-----Evaluation is finished------')
early_x_data.append(clf_threshold_pre)
early_y_data.append(frames_early_meter.avg)
early_plot.set_xdata(early_x_data)
early_plot.set_ydata(early_y_data)
plt.annotate(f'{levenshtein_accuracies.avg * 100:.2f}',
(clf_threshold_pre, frames_early_meter.avg),
textcoords='offset pixels',
xytext=(10, 10))
plt.gca().relim()
plt.gca().autoscale_view()
plt.pause(0.0001)
plt.draw()
