Open MuhammadAsadJaved opened 5 years ago
there is actually a pre-trained model for pedestrian detection using aerial images. you can find it if you look under the ready to use models on the app.
to run the model on a video, you would have to break the video into frames and get your predictions for each frame separately.
here's a script to run it in real time - https://github.com/NanoNets/real-time-python-nanonets/blob/master/nanonets.py
I have two questions. 1-is it necessary to train new models? why don't you provide pre-trained models? Can anyone share the pre-trained models if anyone has done it? 2-How to run the trained model on video file or on webcam?