detection.py

""" Performs detection using trained model.

CL Args:
  -i Path to input video file.
  -o Path to output video file.
  --model Path to trained model.
  --heat_map Path to heatmap save file.
"""

from nets import SegmentationModel
from skimage.transform import rescale
import numpy as np
import matplotlib.pyplot as plt
import skvideo.io
from skimage.color import rgb2gray
from skimage.util import img_as_ubyte
from keras.utils import multi_gpu_model
from util import get_parser
from scipy.ndimage.measurements import label
import warnings
warnings.filterwarnings("ignore")


def heat_map(frame, m, n):

    i, j = frame.shape

    HM = np.zeros((i // m, j // n))

    for ii in np.arange(0, i, m):
        for jj in np.arange(0, j, n):
            pom = frame[ii:ii + m, jj:jj + n]

            labeled_array, num_labels = label(pom)

            HM[ii // m, jj // n] = num_labels

    return HM


patch_size = 64
m = 64
n = 64

args = get_parser().parse_args()

INPUT_FILE = args.input
outfile = args.output
heat = args.heat_map


metadata = skvideo.io.ffprobe(INPUT_FILE)
rate = metadata['video']['@r_frame_rate']
width = int(metadata['video']['@width'])
height = int(metadata['video']['@height'])
videodata = skvideo.io.FFmpegReader(INPUT_FILE)

writer = skvideo.io.FFmpegWriter(outfile, inputdict={'-r': rate},
                                 outputdict={'-r': rate, '-pix_fmt': 'yuv420p'})

model = SegmentationModel(input_shape=(height, width, 5))

try:
    model = multi_gpu_model(model)
except Exception:
    pass

model.load_weights(args.model)

X = np.zeros((1, height, width, 5))

X[0, :, :, 0] = rgb2gray(next(videodata.nextFrame()))
X[0, :, :, 0] -= 0.5
X[0, :, :, 1] = rgb2gray(next(videodata.nextFrame()))
X[0, :, :, 1] -= 0.5
X[0, :, :, 2] = rgb2gray(next(videodata.nextFrame()))
X[0, :, :, 2] -= 0.5
X[0, :, :, 3] = rgb2gray(next(videodata.nextFrame()))
X[0, :, :, 3] -= 0.5

writer.writeFrame(img_as_ubyte(np.zeros((height, width))))
writer.writeFrame(img_as_ubyte(np.zeros((height, width))))

try:
    for k, f in enumerate(videodata.nextFrame()):
        print(k)
        X[0, :, :, 4] = rgb2gray(f)
        X[0, :, :, 4] -= 0.5

        pred = model.predict(X)
        ff = rescale(pred[0, :, :, 0], 2, anti_aliasing=False)

        writer.writeFrame(img_as_ubyte(ff))

        frame_1 = ff[0:(height // patch_size) * patch_size, 0:(width // patch_size) * patch_size] > 0.15

        HM = heat_map(frame_1, m, n)

        if k == 0:
            mapa = HM
        else:
            mapa = mapa + HM

        X[0, :, :, 0] = X[0, :, :, 1].copy()
        X[0, :, :, 1] = X[0, :, :, 2].copy()
        X[0, :, :, 2] = X[0, :, :, 3].copy()
        X[0, :, :, 3] = X[0, :, :, 4].copy()
except RuntimeError:
    a = 0

plt.figure()
plt.imshow(mapa, cmap='hot')
plt.colorbar()
plt.title("Heatmap of input video on patch size " + str(m) + "x" + str(n) + ".")
plt.savefig(heat + str(m) + 'x' + str(n) + '.pdf', bbox_inches='tight')

np.save(heat + str(m) + 'x' + str(n), mapa)

videodata.close()
writer.close()