corresponding proxy, synthetic proxy and embedding

[pribadihcr]

Hi, According to figure 3 in the README.

• Proxy (train) is defined as "self.proxy" in the script https://github.com/navervision/proxy-synthesis/blob/master/loss/proxy_losses.py#L57
• Synthetic proxy (train) and Synthetic embedding (train) are return values of https://github.com/navervision/proxy-synthesis/blob/master/loss/proxy_losses.py#L14 (input_l2, proxy_l2)

Is it correct?

[geonm]

Hi @pribadihcr

Proxy (train) is defined as "self.proxy" in the script https://github.com/navervision/proxy-synthesis/blob/master/loss/proxy_losses.py#L57

• Yes. Original proxy is defined as "self.proxy"

Synthetic proxy (train) and Synthetic embedding (train) are return values of https://github.com/navervision/proxy-synthesis/blob/master/loss/proxy_losses.py#L14 (input_l2, proxy_l2)

• The return values of proxy_synthesis function contain original proxy and embedding.
  Please refer to https://github.com/navervision/proxy-synthesis/blob/master/loss/proxy_losses.py#L23-L25

[pribadihcr]

Hi @geonm , thanks for the replay. So, How to output the synthetic proxy and synthetic embedding according to figure 3 in the README ?.