I am trying to evaluate a pretranined model created by EpicKitchens. But I stuck at an error when I create my own data.
Link of pretrained model is indicated below.
class OwnDataset(Dataset):
running_loss = 0.0
running_corrects = 0
def __init__(self, image_paths,transform, length):
self.image_paths = image_paths
self.transform = transform
self.length = length
def __getitem__(self, index):
dimensions =[]
image_path_at_index = self.image_paths[index]
video_name = image_path_at_index.split('/')[-2] # P01_01
frame_name = image_path_at_index.split('/')[-1] # frame_0000000001.jpg
frame_number = int(frame_name.split('_')[-1][:-4]) # 1
# print(frame_number)
#frame_number = 9
selected_row = df_train_action_labels.query(f"video_id == \"{video_name}\" and {frame_number}>= start_frame and {frame_number}<=stop_frame")
selected_row = selected_row.drop(columns=['uid','participant_id','start_timestamp', 'stop_timestamp', 'start_frame', 'stop_frame', 'all_nouns', 'all_noun_classes'])
row_list = []
for index, rows in selected_row.iterrows():
my_list =[rows.video_id ,rows.verb, rows.verb_class, rows.noun, rows.noun_class]
row_list.append(my_list)
# print(row_list)
# print("image_path_at_index: ", image_path_at_index) image_path_at_index: /content/drive/My Drive/Videos/train/P01_01/frame_0000000001.jpg
image = Image.open(image_path_at_index)
if self.transform:
image = self.transform(image)
# train_video_names
# df_train_action_labels[df_train_action_labels.video_id == "P01_01"]
x = image
y = row_list
return x, y
def __len__(self):
return self.length
df_train_action_labels = pd.read_csv(annotations_dir +'/EPIC_train_action_labels.csv')
class_paths_train = [d.path for d in os.scandir(train_dir) if d.is_dir]
whole_train_image_paths = []
train_video_names = [path.split('/')[-1] for path in class_paths_train]
for c, class_path in enumerate(class_paths_train):
paths = sorted(glob.glob(os.path.join(class_path, '*.jpg')))
whole_train_image_paths.extend(paths)
# print(whole_train_image_paths) ['/content/drive/My Drive/Videos/test/P01_12/0000000001.jpg', '/content/drive/My Drive/Videos/test/P01_12/0000000031.jpg', '/....]
transform = transforms.Compose([
transforms.Resize(size=(224, 224)),
transforms.ToTensor(),
transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
dataset_train = OwnDataset(
image_paths = whole_train_image_paths,
transform = transform,
length = len(whole_train_image_paths)
)
loader_train = DataLoader(
dataset_train,
batch_size=1,
num_workers=1
)
for i, data in enumerate(loader_train):
inputs,labels = data
output = tsn.features(inputs)
verb_logits, noun_logits = tsn.logits(features)
if isinstance(output, tuple):
print(f"Outputsssss shape: {[o.shape for o in output]}")
else:
print(f"Output shape: {output.shape}")
And I get an error
RuntimeError: shape '[-1, 8, 125]' is invalid for input of size 125
I can’t find its solution what should I do?
Here is an example of correct input shape
batch_size = 1
segment_count = 8
snippet_length = 1 # Number of frames composing the snippet, 1 for RGB, 5 for optical flow
snippet_channels = 3 # Number of channels in a frame, 3 for RGB, 2 for optical flow
height, width = 224, 224
inputs = torch.randn(
[batch_size, segment_count, snippet_length, snippet_channels, height, width]
)
# The segment and snippet length and channel dimensions are collapsed into the channel
# dimension
# Input shape: N x TC x H x W
print("b1" , inputs.shape #[1, 8, 1, 3 224, 224]
inputs = inputs.reshape((batch_size, -1, height, width))
print( inputs.shape) #[1, 24, 224, 224]
for model in [tsn]:
# You can get features out of the models
features = model.features(inputs)
# and then classify those features
verb_logits, noun_logits = model.logits(features)
# or just call the object to classify inputs in a single forward pass
verb_logits, noun_logits = model(inputs)
print(verb_logits.shape, noun_logits.shape)
