sns.set_context("talk")
clip_attrs = pd.read_csv("clip_attrs.csv")
videos = np.sort(glob.glob("*.mp4"))
print(videos)
# Process each video using our detector.
# In[ ]:
from feat import Detector
detector = Detector(au_model = "rf", emotion_model = "resmasknet")
for video in videos:
detector.detect_video(video, outputFname = video.replace(".mp4", ".csv"))
# In[9]:
from feat.utils import read_feat
import pandas as pd
for ix ,video in enumerate(videos):
outputF = video.replace(".mp4", ".csv")
if ix == 0:
fex = read_feat(outputF)
else:
fex = pd.concat([fex, read_feat(outputF)])
fex = fex.dropna()
# Find the file you want to process. from feat.tests.utils import get_test_data_path import os, glob test_data_dir = get_test_data_path() test_video = os.path.join(test_data_dir, "WolfgangLanger_Pexels.mp4") # Show video from IPython.display import Video Video(test_video, embed=True) # Let's predict facial expressions from the video using the `detect_video()` method. # In[22]: video_prediction = detector.detect_video(test_video, skip_frames=24) video_prediction.head() # You can also plot the detection results from a video. The frames are not extracted from the video (that will result in thousands of images) so the visualization only shows the detected face without the underlying image. # # The video has 24 fps and the actress show sadness around the 0:02, and happiness at 0:14 seconds. # In[23]: video_prediction.loc[[48]].plot_detections() # In[24]: video_prediction.loc[[17 * 24]].plot_detections(pose=True) # We can also leverage existing pandas plotting functions to show how emotions unfold over time. We can clearly see how her emotions change from sadness to happiness.