Viewing profile — rdabane

What is commonly used tool for 'fast' annotation of video frames with bounding boxes ( e.g. Face detection)?

I'm looking for annotated training/test video data sequences for driver monitoring(Driver drowsiness etc) application development.

I can attest to this fact ... I was a tutor to freshmans during my grad studies. The kids couldn't handle simple sum of fractions without a calculator. Forget about sum fractions c…

Thanks. Would you happen to know where these firms source the human labor from?

I'm looking for outsourcing firms to get hours of video data annotated .. on the topic, are there any good open source tools for video annotations ?

I'm highly interested in learning about genomics. What's the best way for an electrical engineer(signal processing,information theory,graph theory) to start gaining more insights i…

I'm wondering in both classical and deep learning approaches if the input is sharpened images, does it affect the quality of detection?

Smallest size=memory required for intermediate + Network

Basically I need to understand if there is dependency to store multiple feature maps at any point in the forward pass?

Assuming I've a MxN input image, I need to estimate minimum memory required for inference. I'm ok with any kind of reordering of computations.

Why most of CNN papers use 224x224 or 223x223 size images? Why not use more standard like 640x480?

Yup, like a near collision scenario to teach it to drive off the road instead of head on collision ...

It should never fail since any failure could potentially create a fatal scenario. People usually accept fatalities because of human error but they won't accept death because of alg…

Its amazing how the characteristics defined in ( https://en.wikipedia.org/wiki/Contextual_learning ) apply directly to this scenario. (Learning from a master )

Thanks.

Thanks for the reference.

Wondering if they have to create near collision scenarios in order to train the network :-)

Yeah,make sense. But this is the first example I've seen where the network is trained only with driver's input. Has anyone seen this kind of approach before?