o
V hS @ s d dl Z d dlZd dlmZ d dlmZmZmZmZm Z m
Z
m Z d dl m
Z
ddlmZmZ ddlmZ ddlmZ G d d
d
eZdS ) N)Path)AnyCallableDictListOptionalTupleUnion)Tensor ) find_classes make_dataset)
VideoClips)
VisionDatasetc ! s e Zd ZdZdZdddZdZdZ
d)d ee e
f d
e de de de e de de
de e de ee ef de de de de de de ddf fd d
Zedee ef fd d Zd ee d!e de de
dee f
d"d#Zde fd$d%Zd&e deeee f fd'd(Z ZS )*HMDB51a
`HMDB51 `_
dataset.
HMDB51 is an action recognition video dataset.
This dataset consider every video as a collection of video clips of fixed size, specified
by ``frames_per_clip``, where the step in frames between each clip is given by
``step_between_clips``.
To give an example, for 2 videos with 10 and 15 frames respectively, if ``frames_per_clip=5``
and ``step_between_clips=5``, the dataset size will be (2 + 3) = 5, where the first two
elements will come from video 1, and the next three elements from video 2.
Note that we drop clips which do not have exactly ``frames_per_clip`` elements, so not all
frames in a video might be present.
Internally, it uses a VideoClips object to handle clip creation.
Args:
root (str or ``pathlib.Path``): Root directory of the HMDB51 Dataset.
annotation_path (str): Path to the folder containing the split files.
frames_per_clip (int): Number of frames in a clip.
step_between_clips (int): Number of frames between each clip.
fold (int, optional): Which fold to use. Should be between 1 and 3.
train (bool, optional): If ``True``, creates a dataset from the train split,
otherwise from the ``test`` split.
transform (callable, optional): A function/transform that takes in a TxHxWxC video
and returns a transformed version.
output_format (str, optional): The format of the output video tensors (before transforms).
Can be either "THWC" (default) or "TCHW".
Returns:
tuple: A 3-tuple with the following entries:
- video (Tensor[T, H, W, C] or Tensor[T, C, H, W]): The `T` video frames
- audio(Tensor[K, L]): the audio frames, where `K` is the number of channels
and `L` is the number of points
- label (int): class of the video clip
zJhttps://serre-lab.clps.brown.edu/wp-content/uploads/2013/10/hmdb51_org.rarzQhttps://serre-lab.clps.brown.edu/wp-content/uploads/2013/10/test_train_splits.rar 15e67781e70dcfbdce2d7dbb9b3344b5)urlmd5r NTr THWCrootannotation_pathframes_per_clipstep_between_clips
frame_ratefoldtrain transform_precomputed_metadata num_workers _video_width
_video_height_video_min_dimension_audio_samples
output_formatreturnc
s t | |dvrtd| d}t| j\| _}t| j||| _dd | jD }t||||| |
| | |
||d }|| _ || _
|| _ | ||||| _
|| j
| _|| _d S )N)r r z$fold should be between 1 and 3, got )avic S s g | ]\}}|qS r( ).0path_r( r( O/var/www/vscode/kcb/lib/python3.10/site-packages/torchvision/datasets/hmdb51.py
[ s z#HMDB51.__init__..)r r r! r" r# r$ )super__init__
ValueErrorr r classesr
samplesr full_video_clipsr r _select_foldindicessubset video_clipsr )selfr r r r r r r r r r r r! r" r# r$
extensions class_to_idx video_pathsr7 __class__r( r, r/ = s<
zHMDB51.__init__c C s | j jS N)r3 metadatar8 r( r( r, r? s s zHMDB51.metadata
video_listannotations_dirc C s |r| j n| j}d| d}tj||}t|}t } |D ]2}
t|
} | } W d n1 s4w Y | D ]}
|
\}}t
|}||krP| | q;q g }t |D ]\}}tj
|| v ri|| qX|S )Nz *test_splitz.txt) TRAIN_TAGTEST_TAGosr* joinglobsetopen readlinessplitintadd enumeratebasenameappend)r8 rA rB r r
target_tagsplit_pattern_namesplit_pattern_pathannotation_pathsselected_filesfilepathfidlineslinevideo_filename
tag_stringtagr5 video_index
video_pathr( r( r, r4 w s,
zHMDB51._select_foldc C s
| j S r> )r7 num_clipsr@ r( r( r, __len__ s
zHMDB51.__len__idxc C sJ | j |\}}}}| j| }| j| \}}| jd ur | |}|||fS r> )r7 get_clipr5 r2 r )r8 ra videoaudior+ video_idx sample_index class_indexr( r( r, __getitem__ s
zHMDB51.__getitem__) r Nr TNNr r r r r r )__name__
__module__ __qualname____doc__data_urlsplitsrC rD r strr rL r boolr r r r/ propertyr? r r4 r` r r
rh
__classcell__r( r( r< r, r
st '
6&$r )rG rE pathlibr typingr r r r r r r torchr
folderr r
video_utilsr visionr r r( r( r( r, s $