o پi@s|dZddlmZddlZddlZddlZddlmZddl m Z ddl m Z m Z ddlm Zdd lmZGd d d eZdS) z AV Reader.)absolute_importN) VideoReader) AudioReader)cpugpu)ndarray) bridge_outc@sTeZdZdZedddddddfddZdd Zd d Zd d ZddZ ddZ dS)AVReaderapIndividual audio video reader with convenient indexing function. Parameters ---------- uri: str Path of file. ctx: decord.Context The context to decode the file, can be decord.cpu() or decord.gpu(). sample_rate: int, default is -1 Desired output sample rate of the audio, unchanged if `-1` is specified. mono: bool, default is True Desired output channel layout of the audio. `True` is mono layout. `False` is unchanged. width : int, default is -1 Desired output width of the video, unchanged if `-1` is specified. height : int, default is -1 Desired output height of the video, unchanged if `-1` is specified. num_threads : int, default is 0 Number of decoding thread, auto if `0` is specified. fault_tol : int, default is -1 The threshold of corupted and recovered frames. This is to prevent silent fault tolerance when for example 50% frames of a video cannot be decoded and duplicate frames are returned. You may find the fault tolerant feature sweet in many cases, but not for training models. Say `N = # recovered frames` If `fault_tol` < 0, nothing will happen. If 0 < `fault_tol` < 1.0, if N > `fault_tol * len(video)`, raise `DECORDLimitReachedError`. If 1 < `fault_tol`, if N > `fault_tol`, raise `DECORDLimitReachedError`. riDTc CsFt|||||_|jt|dr|dt|||||||_dS)Nreadr)r_AVReader__audio_reader add_paddinghasattrseekr_AVReader__video_reader) selfurictx sample_ratemonowidthheight num_threads fault_tolrD/home/ubuntu/.local/lib/python3.10/site-packages/decord/av_reader.py__init__,s    zAVReader.__init__cCs t|jS)zGet length of the video. Note that sometimes FFMPEG reports inaccurate number of frames, we always follow what FFMPEG reports. Returns ------- int The number of frames in the video file. )lenr)rrrr__len__3s zAVReader.__len__cCs|jdur |jdus Jt|tr|t|t|jS|dkr)|t|j7}|t|jks4|dkr?td |t|j|j |\}}|j |}|j |}|j|||j|fS)aGet audio samples and video frame at `idx`. Parameters ---------- idx : int or slice The frame index, can be negative which means it will index backwards, or slice of frame indices. Returns ------- (ndarray/list of ndarray, ndarray) First element is samples of shape CxS or a list of length N containing samples of shape CxS, where N is the number of frames, C is the number of channels, S is the number of samples of the corresponding frame. Second element is Frame of shape HxWx3 or batch of image frames with shape NxHxWx3, where N is the length of the slice. NrzIndex: {} out of bound: {}) rr isinstanceslice get_batchrangeindicesr IndexErrorformatget_frame_timestamp_time_to_sample)ridxaudio_start_idx audio_end_idxrrr __getitem__=s   zAVReader.__getitem__c Cs|jdur |jdus J||}g}d}d}t|D]/}|j|\}}|r0||dkr0|}n|j|}|j|} ||j|| |}| }q||j|fS)aGet entire batch of audio samples and video frames. Parameters ---------- indices : list of integers A list of frame indices. If negative indices detected, the indices will be indexed from backward Returns ------- (list of ndarray, ndarray) First element is a list of length N containing samples of shape CxS, where N is the number of frames, C is the number of channels, S is the number of samples of the corresponding frame. Second element is Frame of shape HxWx3 or batch of image frames with shape NxHxWx3, where N is the length of the slice. Nr)rr _validate_indiceslistr'r(appendr") rr$ audio_arrprev_video_idxprev_audio_end_idxr)frame_start_timeframe_end_timer*r+rrrr"\s    zAVReader.get_batchcCstj|jddfdd}t|D]&}|j|\}}|j|}|j|}tj||j|| fdd}qt t ||j |fS)Nrfloat32)shapedtyper)axis)npemptyr r6r.rr'r( concatenateasnumpyr _ndarrayr")rslr0r)r*r+rrr _get_slices   "zAVReader._get_slicecCs|jdur |jdus Jtj|tjd}||dkt|j7<|dks7td||dkt|j|t|jksNtd||t|jk|S)zTValidate int64 integers and convert negative integers to positive by backward searchN)r7rzInvalid negative indices: {}zOut of bound indices: {}) rr r9r>int64rallr%r&)rr$rrrr-s zAVReader._validate_indicesN) __name__ __module__ __qualname____doc__rrrr,r"r@r-rrrrr s % r )rF __future__rctypesnumpyr9math video_readerr audio_readerrrrrr=bridger objectr rrrrs