10:42:30	 From  Louis Moreau (Edge Impulse) : Happy to take your question either here or live ?
10:43:27	 From  Al Anway : does the face recognition work without modification if the image is rotated, say, upside down?
10:45:46	 From  Gillian Minnehan : You mentioned that objects close together and overlapped are not detected well. Does Edge impulse have ideas or plans on how to fix this?
10:45:56	 From  Gillian Minnehan : Or improve at least
10:46:55	 From  Tiago Butzke dos Santos : Does the flash image size increase substantially when training for more than one class of object? What's the impact in performance when detecting more than one class?
10:47:46	 From  Gillian Minnehan : Thank you!
10:48:51	 From  Raul Gotor : Can we train the model with a set of frames, (i.e: short video) in such a way that we can distinguish a person walking from one standing still?
10:48:56	 From  Tiago Butzke dos Santos : Thanks!
10:49:11	 From  Al Anway : if you wanted to count people in a crowd, or caribou, or elephants, where the number is very large, how would you go about it?
10:50:37	 From  Raul Gotor : Ok, many thanks!
10:52:34	 From  Gillian Minnehan : The business model of having tools to make ML easier for developers is pretty popular. Lots of consulting firms and companies like C3AI, Palantir, etc. What makes Edge impulse different? Is it simply the specific focus on IoT devices?
10:54:06	 From  Gillian Minnehan : Okay that makes sense. Thank you!
10:54:29	 From  Tobias Lifschitz : Hi Louis, thanks for the talk! Not quite related to the main topic, but when dealing with an unlabled time series dataset (vibration monitoring), does edge impulse offer tools for anomaly detection or unsupervised learning models?
10:57:07	 From  Tobias Lifschitz : Thats great! I'll look into it, thanks
10:57:44	 From  Al Anway : great talk!
10:59:34	 From  Tiago Butzke dos Santos : Awesome work!
10:59:34	 From  Gillian Minnehan : Thank you Louis!