redesign

[ggsato]

Goal: clear concept

• very strong value proposition ==> 2 stages analysis(flow and scene) ===> collaborative tracking with optical flow and detection ===> weighed graph based flow analysis ==> flexible events for rich GUI apps
• no maybe features
• sort out anything ugly

[ggsato]

As long as it is based on deep learning, an automatic learning is a critical key.

Detection can not be trained automatically.

Re-identification is possible through tracking sequences. But to do this, a tracking should produce a reasonable rectangles, not too small nor too large. Thus, a smooth tracking is a key feature to achieve this.