Open ubergeekNZ opened 10 months ago
It looks like you are feeding a single image into the CNN at a time. The CNN is designed to calculate the optical flow between a pair of frames, and must be fed both frames each time. Unlike the models used during training, the TensorFlow Lite model that you are using here actually has two separate inputs for the frames. You can easily visualize this with a tool like netron.
Based on @ubergeekNZ 's code provided here I made a google colab notebook that uses your webcam to test out nanoflownet.
Hi,
I was wondering if I am using the tflite model correctly and visualizing using the flow_vis function. I don't seem to be able to get similar images to the paper.
Paul