Video Pipeline Reading Labelled Videos from a Directory¶
In this example, we will go through the creation of a pipeline using the readers.video operator to read videos along with their labels. The pipeline will return a pair of outputs: a batch of sequences and respective labels.
For more information on the readers.video parameters, please look at the documentation.
First let’s start with the imports:
import os import numpy as np from nvidia.dali import pipeline_def import nvidia.dali.fn as fn import nvidia.dali.types as types
We need some video containers to process. We can use Sintel trailer, which is an mp4 container containing an h.264 video and distributed under the Create Common license. We’ve split it into 5 second clips and divided the clips into labelled groups. This can be done easily with the
ffmpeg standalone tool.
Then we can set the parameters that will be used in the pipeline. The
count parameter will define how many frames we want in each sequence sample.
We can replace
video_directory with any other directory containing labelled subdirectories and video container files recognized by FFmpeg.
batch_size=2 sequence_length=8 initial_prefetch_size=11 video_directory = os.path.join(os.environ['DALI_EXTRA_PATH'], "db", "video", "sintel", "labelled_videos") shuffle=True n_iter=6
DALI_EXTRA_PATH environment variable should point to the place where data from DALI extra repository is downloaded. Please make sure that the proper release tag is checked out.
Running the Pipeline¶
We can then define a minimal Pipeline that will output directly the readers.Video outputs:
@pipeline_def def video_pipe(file_root): video, labels = fn.readers.video(device="gpu", file_root=file_root, sequence_length=sequence_length, random_shuffle=True, initial_fill=initial_prefetch_size) return video, labels
Caution: One important here is tuning
initial_fill, that correspond to the Loader prefetch buffer intial size. Since this buffer will be filled of
initial_fill sequences, the total number of frames can be really huge! So set it consequently to not OOM during training.
Let’s try to build and run a
video_pipe instance on device 0 that will output
batch_size sequences of
count frames and
batch_size labels at each iteration.
pipe = video_pipe(batch_size=batch_size, num_threads=2, device_id=0, file_root=video_directory, seed=12345) pipe.build() for i in range(n_iter): sequences_out, labels = pipe.run() sequences_out = sequences_out.as_cpu().as_array() labels = labels.as_cpu().as_array() print(sequences_out.shape) print(labels.shape)
(2, 8, 720, 1280, 3) (2, 1) (2, 8, 720, 1280, 3) (2, 1) (2, 8, 720, 1280, 3) (2, 1) (2, 8, 720, 1280, 3) (2, 1) (2, 8, 720, 1280, 3) (2, 1) (2, 8, 720, 1280, 3) (2, 1)
Visualizing the Results¶
The previous iterations seems to have the yield batches of the expected shape. But let’s visualize the results to be
sequences_out, labels = pipe.run() sequences_out = sequences_out.as_cpu().as_array() labels = labels.as_cpu().as_array()
We will use matplotlib to display the frames we obtained in the last batch.
%matplotlib inline from matplotlib import pyplot as plt import matplotlib.gridspec as gridspec
def show_sequence(sequence, label): columns = 4 rows = (sequence_length + 1) // (columns) fig = plt.figure(figsize = (32,(16 // columns) * rows)) gs = gridspec.GridSpec(rows, columns) for j in range(rows*columns): plt.subplot(gs[j]) plt.axis("off") plt.suptitle("label " + str(label), fontsize=30) plt.imshow(sequence[j])
And now let’s generate 5 batches of sequence, label pairs:
ITER = 5 for i in range(ITER): sequences_out, labels = pipe.run() sequences_out = sequences_out.as_cpu().as_array() labels = labels.as_cpu().as_array() show_sequence(sequences_out, labels)