Fancy video but seems to be made for investors as target audience.
However, two statements sound really big-headed and disrespectful: - "Today, conventional image sensors are useful for one thing: to take wonderful pictures." - "This approach is fundamentally wrong."
Yes terific video, it is a difficult task to explain in simple terms a technology based company. I would not say the "fundamentally wrong" statement is disrespectful as many of us know, Chronocam does have a point in its approach. The only question to ask a technology startup is whether it can deliver value and capture this value, this requires a lot of smart decisions, not just marketing...
Even this video is worse from technical perspective, it raises interesting questions: Assuming we can get meaning from the picture by throwing away all pixel information and leaving only description of what is going on and how it looks like on the picture. Then how we can transfer this meaning to human brain in more efficient way than the same human will look at the original pixel based picture? What level of detail in description can be achieved and what level of confidence? etc.
I always tell my customers that the best marketing argument is : "We are mimicing the human eye !". Because everybody knows that the human eye is a great image sensor and its working principle is still not completely understood. So if you can mimic the human eye, you must do a great job. Apparently Chronocam knows this story as well ....
AER sensor, like human eye, have some good advantage with a precise timing event, but electronic implementation needs wide area pixel (around 20x20 µm²) and limits the sensor size to VGA organization (0,3 Mp). Some other company have a same project with 9x9 µm² in VGA. It is far from actual 10 Mp organization or more. With TSV in 3D implementation, this process will be on the god way. Perception is not understanding, Cortex have a representation of global blobs like hue and saturation, dynamic motion in direction and velocity (AER sensor in best position), and structural forms as oriented borders, curvatures and textures. Theses 3 representations are done on the cortex and needs to be implemented to understand AER data, it's an other challenge... P. Pirim
An Invisage-style Web video company :)-
ReplyDeleteFancy video but seems to be made for investors as target audience.
ReplyDeleteHowever, two statements sound really big-headed and disrespectful:
- "Today, conventional image sensors are useful for one thing: to take wonderful pictures."
- "This approach is fundamentally wrong."
Terrific marketing video... Chronocam definitely needs to hire a good Marketing chief
ReplyDeleteYes terific video, it is a difficult task to explain in simple terms a technology based company.
ReplyDeleteI would not say the "fundamentally wrong" statement is disrespectful as many of us know, Chronocam does have a point in its approach.
The only question to ask a technology startup is whether it can deliver value and capture this value, this requires a lot of smart decisions, not just marketing...
Even this video is worse from technical perspective, it raises interesting questions: Assuming we can get meaning from the picture by throwing away all pixel information and leaving only description of what is going on and how it looks like on the picture. Then how we can transfer this meaning to human brain in more efficient way than the same human will look at the original pixel based picture? What level of detail in description can be achieved and what level of confidence? etc.
ReplyDeleteChronocam aims machine vision, the human brain is not the target...
DeleteSexy voice. However, not really appropriate for the technical content ;-) BTW, I think their approach is fundamentally wrong for must use cases...
ReplyDeleteI always tell my customers that the best marketing argument is : "We are mimicing the human eye !". Because everybody knows that the human eye is a great image sensor and its working principle is still not completely understood. So if you can mimic the human eye, you must do a great job. Apparently Chronocam knows this story as well ....
ReplyDeleteAER sensor, like human eye, have some good advantage with a precise timing event, but electronic implementation needs wide area pixel (around 20x20 µm²) and limits the sensor size to VGA organization (0,3 Mp). Some other company have a same project with 9x9 µm² in VGA. It is far from actual 10 Mp organization or more. With TSV in 3D implementation, this process will be on the god way.
ReplyDeletePerception is not understanding, Cortex have a representation of global blobs like hue and saturation, dynamic motion in direction and velocity (AER sensor in best position), and structural forms as oriented borders, curvatures and textures. Theses 3 representations are done on the cortex and needs to be implemented to understand AER data, it's an other challenge...
P. Pirim