Hi, in ordering to understand a bit better, how the Neural Networks works, I found an Interesting project TensorSpace to visualize different pre-trained deep learning models, including object classification, object detections, image generations etc.

All models in the "Playground" are interactive. We can move the mouse to see the relation lines among layers; we can click the layer aggregation to check feature maps; we can move the camera to view from any direction.

- left click and drag -

change camera angle

- right click and drag -

change camera position

- scroll up and down -

move camera closer or further