ejcgt / attention-target-detection

[CVPR2020] "Detecting Attended Visual Targets in Video"
MIT License
173 stars 48 forks source link

where is demo data #2

Closed Jackwanguestc closed 4 years ago

Jackwanguestc commented 4 years ago

hello , show in line 21 in demo.py, run demo need data/ dir, where to download?

parser.add_argument('--image_dir', type=str, help='images', default='data/demo/frames') parser.add_argument('--head', type=str, help='head bounding boxes', default='data/demo/person1.txt')

oliveiracwb commented 4 years ago

Do you will need to implement your own recognition code. I used face_recorgnition with dlib. Fragile results, unfortunately.

ejcgt commented 4 years ago

Hi @Jackwanguestc sorry for missing the data! They're now added please check out the repo again.

ejcgt commented 4 years ago

@oliveiracwb yeah I am working on a fully automated pipleline, but currently you would have to rely on a detection/recognition module of your choice which may not be super reliable in practice unfortunately.