分析注释 sequence_input_layer.py

最新推荐文章于 2024-06-12 11:33:37 发布

置顶

少年冬郎

最新推荐文章于 2024-06-12 11:33:37 发布

阅读量3.7k

点赞数

分类专栏： group activity 文章标签： LSTM gruop activity 视频 sequence_input_layer

本文链接：https://blog.csdn.net/u013110060/article/details/55224409

版权

本文详细分析了sequence_input_layer.py文件，重点探讨其在处理视频数据，尤其是LSTM和group activity任务中的应用。内容涉及如何组织视频数据，包括视频名称列表和按视频划分的数据字典。

摘要由CSDN通过智能技术生成

video_order ： list，存放视频名称，对应视频。

video_dict ：字典，每一个视频下都对应dict里的一项。

#!/usr/bin/env python

#Data layer for video.  Change flow_frames and RGB_frames to be the path to the flow and RGB frames.

import sys
sys.path.append('../../python')
import caffe
import io
from PIL import Image
import matplotlib.pyplot as plt
import numpy as np
import scipy.misc
import time
import pdb
import glob
import pickle as pkl
import random
import h5py
from multiprocessing import Pool
from threading import Thread
import skimage.io
import copy

flow_frames = 'flow_images/'
RGB_frames = 'frames/'
test_frames = 16 
train_frames = 16
test_buffer = 3
train_buffer = 24    # test_frame指的是每一个video提取16帧连续图片，buffer指一次性处理几个video

def processImageCrop(im_info, transformer, flow):
  im_path = im_info[0]
  im_crop = im_info[1] 
  im_reshape = im_info[2]
  im_flip = im_info[3]
  data_in = caffe.io.load_image(im_path)
  if (data_in.shape[0] < im_reshape[0]) | (data_in.shape[1] < im_reshape[1]):
    data_in = caffe.io.resize_image(data_in, im_reshape)
  if im_flip:
    data_in = caffe.io.flip_image(data_in, 1, flow) 
    data_in = data_in[im_crop[0]:im_crop[2], im_crop[1]:im_crop[3], :] 
  processed_image = transformer.preprocess('data_in',data_in)
  return processed_image    #image crop

class ImageProcessorCrop(object):
  def __init__(self, transformer, flow):
    self.transformer = transformer
    self.flow = flow
  def __call__(self, im_info):
    return processImageCrop(im_info, self.transformer, self.flow)

class sequenceGeneratorVideo(object):
  def __init__(self, buffer_size, clip_length, num_videos, video_dict, video_order):
    self.buffer_size = buffer_size
    self.clip_length = clip_length
    self.N = self.buffer_size*self.clip_length
    self.num_videos = num_videos
    self.video_dict = video_dict
    self.video_order = video_order
    self.idx = 0

  def __call__(self):
    label_r = []
    im_paths = []
    im_crop = []
    im_reshape = []  
    im_flip = []
 
    if self.idx + self.buffer_size >= self.num_videos:
      idx_list = range(self.idx, self.num_videos)
      idx_list.extend(range(0, self.buffer_size-(self.num_videos-self.idx))) #short for buffer_size,use the first ,second..complete!
    else:
      idx_list = range(self.idx, self.idx+self.buffer_size)  #once a buffer_size
    # 如果超出了video的数量ÿ